fix(ui): unify heading help, table sizing and navigation contracts

Verified with the coordinated workspace changes by devkit full run
2026-09-08T225814-186389-0000-3e3ed7cd (all seven phases passed).
This shared UI pass does not mark the individual module reviews complete.

.gitignore

@@ -138,6 +138,7 @@ dist
 
 # Local WebUI test/build scratch directories
 .component-test-build/
+.component-test-build-*/
 .file-drop-test-build/
 .module-test-build/
 .policy-test-build/
@@ -149,6 +150,8 @@ webui/.module-test-build/
 webui/.policy-test-build/
 webui/.template-preview-test-build/
 webui/.import-test-build/
+webui/dist-conformance/
+webui/test-results/
 
 # Security audit reports
 audit-reports/

AGENTS.md

@@ -26,18 +26,27 @@ cd /mnt/DATA/git/govoplan-core
 
 For WebUI checks:
 
+```bash
+cd /mnt/DATA/git/govoplan
+./devkit doctor --repo core
+./devkit check --profile ui --repo core --dry-run
+./devkit check --profile ui --repo core
+```
+
+For an individually selected component batch (one compilation):
+
 ```bash
 cd /mnt/DATA/git/govoplan-core/webui
-PATH=/mnt/DATA/git/govoplan-core/webui/node_modules/.bin:/home/zemion/.nvm/versions/node/v22.22.3/bin:$PATH /home/zemion/.nvm/versions/node/v22.22.3/bin/npm run test:mail-components
-PATH=/mnt/DATA/git/govoplan-core/webui/node_modules/.bin:/home/zemion/.nvm/versions/node/v22.22.3/bin:$PATH /home/zemion/.nvm/versions/node/v22.22.3/bin/npm run test:module-capabilities
-PATH=/mnt/DATA/git/govoplan-core/webui/node_modules/.bin:/home/zemion/.nvm/versions/node/v22.22.3/bin:$PATH /home/zemion/.nvm/versions/node/v22.22.3/bin/npm run test:module-permutations
+npm run test:components -- mail-components page-layout
+npm run test:module-capabilities
+npm run test:module-permutations
 ```
 
 Run the consolidated focused check when a change touches module discovery, optional integrations, shared mail components, or mailbox listing:
 
 ```bash
 cd /mnt/DATA/git/govoplan
-tools/checks/check-focused.sh
+./devkit check --profile full
 ```
 
 ## Working Rules

README.md

@@ -33,6 +33,8 @@ Canonical policy documents live in `docs/`:
 - [ACCESS_RBAC_MODEL.md](docs/ACCESS_RBAC_MODEL.md)
 - [GOVERNANCE_MODEL.md](docs/GOVERNANCE_MODEL.md)
 - [MODULE_ARCHITECTURE.md](docs/MODULE_ARCHITECTURE.md)
+- [INTEGRITY_PERFORMANCE_CONTRACT.md](docs/INTEGRITY_PERFORMANCE_CONTRACT.md)
+- [TABULAR_SOURCE_CONTRACT.md](docs/TABULAR_SOURCE_CONTRACT.md)
 - [DEPLOYMENT_OPERATOR_GUIDE.md](docs/DEPLOYMENT_OPERATOR_GUIDE.md)
 - [CODEX_WORKFLOW.md](docs/CODEX_WORKFLOW.md)
 

alembic/dev_versions/a36d8e4f9b12_german_reference_locale.py

@@ -0,0 +1,23 @@
+from __future__ import annotations
+
+from importlib.util import module_from_spec, spec_from_file_location
+from pathlib import Path
+
+
+_path = (
+    Path(__file__).resolve().parents[1]
+    / "versions"
+    / "a36d8e4f9b12_german_reference_locale.py"
+)
+_spec = spec_from_file_location("govoplan_german_reference_locale_migration", _path)
+if _spec is None or _spec.loader is None:
+    raise RuntimeError(f"Unable to load migration implementation from {_path}")
+_module = module_from_spec(_spec)
+_spec.loader.exec_module(_module)
+
+revision = _module.revision
+down_revision = _module.down_revision
+branch_labels = _module.branch_labels
+depends_on = _module.depends_on
+upgrade = _module.upgrade
+downgrade = _module.downgrade

alembic/dev_versions/b47e6f809a13_data_subject_requests.py

@@ -0,0 +1,23 @@
+from __future__ import annotations
+
+from importlib.util import module_from_spec, spec_from_file_location
+from pathlib import Path
+
+
+_path = (
+    Path(__file__).resolve().parents[1]
+    / "versions"
+    / "b47e6f809a13_data_subject_requests.py"
+)
+_spec = spec_from_file_location("govoplan_data_subject_requests_migration", _path)
+if _spec is None or _spec.loader is None:
+    raise RuntimeError(f"Unable to load migration implementation from {_path}")
+_module = module_from_spec(_spec)
+_spec.loader.exec_module(_module)
+
+revision = _module.revision
+down_revision = _module.down_revision
+branch_labels = _module.branch_labels
+depends_on = _module.depends_on
+upgrade = _module.upgrade
+downgrade = _module.downgrade

alembic/dev_versions/c58a2d7e9f10_ownership_decision_history.py

@@ -0,0 +1,21 @@
+"""Development-track wrapper for the ownership history repair."""
+
+from __future__ import annotations
+
+from importlib.util import module_from_spec, spec_from_file_location
+from pathlib import Path
+
+
+_path = Path(__file__).resolve().parents[1] / "versions" / "c58a2d7e9f10_ownership_decision_history.py"
+_spec = spec_from_file_location("govoplan_ownership_decision_history_migration", _path)
+if _spec is None or _spec.loader is None:
+    raise RuntimeError(f"Unable to load migration implementation from {_path}")
+_module = module_from_spec(_spec)
+_spec.loader.exec_module(_module)
+
+revision = _module.revision
+down_revision = _module.down_revision
+branch_labels = _module.branch_labels
+depends_on = _module.depends_on
+upgrade = _module.upgrade
+downgrade = _module.downgrade

alembic/env.py

@@ -26,7 +26,9 @@ from govoplan_core.tenancy.scope import scope_registry
 
 config = context.config
 database_url = config.attributes.get("database_url") or settings.database_url
-config.set_main_option("sqlalchemy.url", database_url)
+# Alembic stores options through ConfigParser: escape only its interpolation
+# syntax so URL-encoded credentials/socket paths reach SQLAlchemy unchanged.
+config.set_main_option("sqlalchemy.url", database_url.replace("%", "%%"))
 
 if config.config_file_name is not None:
     # Migrations can run inside the long-lived application process when module

alembic/versions/a36d8e4f9b12_german_reference_locale.py

@@ -0,0 +1,44 @@
+"""adopt German as the untouched system reference locale
+
+Revision ID: a36d8e4f9b12
+Revises: f25c9d3e7a01
+Create Date: 2026-08-05 00:00:00.000000
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "a36d8e4f9b12"
+down_revision = "f25c9d3e7a01"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    bind = op.get_bind()
+    if "core_system_settings" not in set(sa.inspect(bind).get_table_names()):
+        return
+
+    settings = sa.table(
+        "core_system_settings",
+        sa.column("id", sa.String),
+        sa.column("default_locale", sa.String),
+        sa.column("created_at", sa.DateTime(timezone=True)),
+        sa.column("updated_at", sa.DateTime(timezone=True)),
+    )
+    bind.execute(
+        settings.update()
+        .where(settings.c.id == "global")
+        .where(settings.c.default_locale == "en")
+        .where(settings.c.created_at == settings.c.updated_at)
+        .values(default_locale="de", updated_at=sa.func.now())
+    )
+
+
+def downgrade() -> None:
+    # Locale selection is user-visible state. A downgrade must not overwrite a
+    # German value that may have been selected explicitly after this migration.
+    pass

alembic/versions/b47e6f809a13_data_subject_requests.py

@@ -0,0 +1,77 @@
+"""add governed data-subject request workflow
+
+Revision ID: b47e6f809a13
+Revises: a36d8e4f9b12
+Create Date: 2026-08-07 00:00:00.000000
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "b47e6f809a13"
+down_revision = "a36d8e4f9b12"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    inspector = sa.inspect(op.get_bind())
+    if "core_data_subject_requests" in inspector.get_table_names():
+        return
+    op.create_table(
+        "core_data_subject_requests",
+        sa.Column("id", sa.String(length=36), nullable=False),
+        sa.Column("tenant_id", sa.String(length=36), nullable=False),
+        sa.Column("reference", sa.String(length=120), nullable=False),
+        sa.Column("request_kind", sa.String(length=30), nullable=False),
+        sa.Column("status", sa.String(length=30), nullable=False),
+        sa.Column("subject", sa.JSON(), nullable=False),
+        sa.Column("purpose", sa.String(length=1000), nullable=False),
+        sa.Column("legal_basis", sa.String(length=1000), nullable=True),
+        sa.Column("due_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("requested_by_account_id", sa.String(length=36), nullable=False),
+        sa.Column("search_result", sa.JSON(), nullable=False),
+        sa.Column("erasure_plan", sa.JSON(), nullable=False),
+        sa.Column("execution_result", sa.JSON(), nullable=False),
+        sa.Column("coverage", sa.JSON(), nullable=False),
+        sa.Column("evidence_sha256", sa.String(length=64), nullable=True),
+        sa.Column("resource_revision", sa.Integer(), nullable=False),
+        sa.Column("completed_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("notes", sa.Text(), nullable=True),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_core_data_subject_requests")),
+    )
+    op.create_index(
+        op.f("ix_core_data_subject_requests_tenant_id"),
+        "core_data_subject_requests",
+        ["tenant_id"],
+        unique=False,
+    )
+    op.create_index(
+        op.f("ix_core_data_subject_requests_status"),
+        "core_data_subject_requests",
+        ["status"],
+        unique=False,
+    )
+    op.create_index(
+        op.f("ix_core_data_subject_requests_due_at"),
+        "core_data_subject_requests",
+        ["due_at"],
+        unique=False,
+    )
+    op.create_index(
+        "ix_core_data_subject_requests_tenant_status",
+        "core_data_subject_requests",
+        ["tenant_id", "status"],
+        unique=False,
+    )
+
+
+def downgrade() -> None:
+    inspector = sa.inspect(op.get_bind())
+    if "core_data_subject_requests" in inspector.get_table_names():
+        op.drop_table("core_data_subject_requests")

alembic/versions/c58a2d7e9f10_ownership_decision_history.py

@@ -0,0 +1,37 @@
+"""repair decision history on previously upgraded ownership tables
+
+Revision ID: c58a2d7e9f10
+Revises: b47e6f809a13
+Create Date: 2026-09-07
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "c58a2d7e9f10"
+down_revision = "b47e6f809a13"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # The original ownership migration gained this column after some databases
+    # had already applied it. create_all/checkfirst cannot upgrade those tables.
+    # Fresh installations already have it; never replace their audit evidence.
+    columns = {column["name"] for column in sa.inspect(op.get_bind()).get_columns(
+        "core_ownership_transfers"
+    )}
+    if "decisions" not in columns:
+        op.add_column(
+            "core_ownership_transfers",
+            sa.Column("decisions", sa.JSON(), nullable=False, server_default=sa.text("'[]'")),
+        )
+
+
+def downgrade() -> None:
+    # Older installations and fresh installations at the preceding revision
+    # differ. Keep the additive column and any subsequently recorded evidence.
+    pass

docs/ACCESS_RBAC_MODEL.md

@@ -142,6 +142,7 @@ system:tenants:read
 system:tenants:create
 system:tenants:update
 system:tenants:suspend
+system:tenants:erase
 
 system:accounts:read
 system:accounts:create

docs/API_CLIENT_CACHE_CONTRACT.md

@@ -0,0 +1,35 @@
+# Shared API client cache and authority boundaries
+
+All optional WebUI modules use the Core API client. Its bounded in-memory caches
+are an optimization, never an authorization mechanism. The backend must check
+the current principal, tenant, and permissions even for conditional GETs.
+
+- Identical simultaneous safe requests can share one network request. Requests
+  with caller-owned cancellation are independent.
+- Responses allowing reuse have at most a 750 ms recent-response window.
+  `no-store` and `Vary: *` responses are not retained. `no-cache` and zero-age
+  responses require a server check; permitted ETags retain conditional GET
+  support without bypassing authorization. This follows the relevant
+  [HTTP cache-control semantics](https://www.rfc-editor.org/rfc/rfc9111.html#section-5.2.2).
+- Explicit `cache: "no-store"`, `"reload"`, or `"no-cache"` reads bypass older
+  response data and supersede older requests for that resource. Reload is not a
+  mutation. Owning read helpers must pass these options through pagination.
+- Writes invalidate caches before execution and again on settlement, including
+  failures whose server outcome may be uncertain. Reads started before or
+  during the write cannot seed reusable data after it finishes.
+- The shell calls `clearApiReadCache()` before explicit auth updates and when
+  refreshing authoritative session data. API-settings changes, clearing the
+  token, authentication expiry, and changes to the paired session/CSRF cookie
+  also invalidate both stored and in-flight reuse. Cookie observation also
+  covers sign-in/out in another tab; it does not read the HttpOnly session token.
+- Interactive sign-in and sign-out clear a previously saved automation key.
+  Explicit key-based connection settings still select the key's identity;
+  profile-only updates preserve settings identity to avoid reload loops.
+- Expired responses from superseded reads or downloads do not trigger a login
+  prompt in a newer session.
+- Every completion (including 304) must still own its cache slot and generation
+  before storing anything. An old caller may receive its own result, so feature
+  components must continue guarding displayed state against obsolete requests.
+
+Regression coverage: `npm run test:api-client-cache` uses the real client and
+isolated network fixtures. No live API or account data is involved.

docs/BOUNDED_PROCESS_CONTRACT.md

@@ -0,0 +1,81 @@
+# Disposable resource-bounded operations
+
+`security.bounded_process.run_bounded_operation` runs a trusted, importable,
+module-level `bytes -> bytes` function in a fresh interpreter. Core owns the
+process lifecycle, not the business parser. Owners retain authorization,
+sessions, provider reads, idempotency and persistence in the parent and pass
+only explicit bounded data. Never accept the operation, module or source path
+from a client. Use `security.worker_payload` for typed values; it does not use
+pickle, arbitrary constructors or JSON object hooks.
+
+The runner requires POSIX process groups, `waitid(WNOWAIT)` and resource limits.
+Unsupported controls fail closed; there is no in-process fallback. The child
+uses `-I -B`, a fixed minimal environment, `/` as working directory, closed
+inherited descriptors and a new process session. Limits are installed before
+the owning module is imported. Installed dependencies must support isolated
+Python imports; development `PYTHONPATH` alone is insufficient.
+
+`ProcessLimits` specifies wall-clock seconds (including child startup), CPU
+seconds, virtual address space, input/output pipe bytes and maximum regular-file
+size. Defaults are 10 seconds wall/CPU, 256 MiB address space, 8 MiB input and
+output, and no regular-file output. Wall/CPU limits are at most 600 seconds,
+memory 64 MiB–8 GiB, pipe limits 1 byte–256 MiB, and file size 0–2 GiB. Owners
+must document their tighter functional limits; a transport cap does not replace
+row, archive expansion, item-count or artifact limits.
+
+Admission is non-queuing. `GOVOPLAN_ISOLATED_PROCESS_CONCURRENCY` defaults to 1
+(range 1–16) and applies across these operations **within each API/worker
+process**. Multiply capacity and memory budgets by the number of API/worker
+processes when sizing an installation. This is not a fleet-wide semaphore,
+cgroup quota, filesystem/network sandbox or permission to run arbitrary code.
+`RLIMIT_FSIZE` is per file, not a total disk quota. Owners creating staged files
+must enforce cumulative quotas and clean up their own private directories.
+
+Owners preparing bounded local snapshots can enter
+`bounded_operation_admission()` before preparation and pass its token as
+`admission=` to the runner. This reuses shared capacity rather than reserving a
+second slot. Tokens belong to their active context, thread and process; expired,
+cross-thread and overlapping reuse fail. Preparation exceptions release the
+slot without launching a child. Never hold admission while waiting for a user;
+parent-side preparation still requires explicit I/O and byte bounds.
+
+The parent concurrently drains stdout/stderr while writing input. Output is
+bounded during reading, stderr is discarded and capped at 64 KiB, and raw child
+tracebacks are never returned. Every success, exception, timeout, cancellation
+and callback failure kills the owned process group before reaping its leader,
+including descendants which close their inherited pipes. A module-level child
+handler must return bytes; it must not print logs/progress to stdout.
+
+The optional `cancelled` callback runs in the parent at most roughly every
+50 ms while waiting. It must be fast and must not return an awaitable. It may
+also service a module-owned bounded progress protocol; exceptions terminate
+the child and propagate. There is no fabricated progress for killed work.
+`ProcessBudgetError.code` distinguishes busy, cancelled, timeout, CPU, memory,
+input/output limits, unavailable controls and worker failure. Owners map these
+to their existing structured diagnostics and recovery semantics.
+
+The private typed-data codec supports null, booleans, strings, bytes, integers,
+floats, Decimal, UUID, date/time/datetime, lists, tuples and string-keyed maps.
+It rejects unsupported objects, malformed/trailing bytes, duplicate keys,
+excess depth (64) and node counts (1,000,000). Operation DTOs remain owner
+contracts and require owner validation. Do not persist this private wire format
+or use it as a public API.
+
+Tests use real child processes for catastrophic regex, memory exhaustion,
+noisy output, exact limits, cancellation, closed-pipe hangs and descendant
+cleanup. These are local process regression tests, not production concurrent
+load certification. Operators still need target Linux/cgroup, cancellation,
+worker-count, memory and disk-quota evidence before raising concurrency.
+
+## Deutsche Betriebszusammenfassung
+
+Rechenintensive, vertrauenswürdige Moduloperationen laufen in einem frischen
+Prozess mit harten Laufzeit-, CPU-, Speicher- und Ausgabegrenzen. Berechtigungen,
+Sitzungen, Zugangsdaten und Datenbankänderungen bleiben im Hauptprozess. Fehlende
+Betriebssystemkontrollen führen zu einer Diagnose, nicht zu ungeschützter
+Ausführung. `GOVOPLAN_ISOLATED_PROCESS_CONCURRENCY` begrenzt die gemeinsame
+Zulassung je API-/Worker-Prozess, standardmäßig auf 1. Mehrere Prozesse haben
+jeweils eigene Grenzen; systemweite Speicher- und Festplattenquoten müssen
+Betreiber zusätzlich konfigurieren und auf der Zielinstallation prüfen. Die
+Schnittstelle ist keine Sandbox für beliebigen Code. Modul-Dokumentation nennt
+die jeweiligen fachlichen Grenzen, Fortschritts- und Wiederholungsregeln.

docs/CONFIGURATION_PACKAGES.md

@@ -157,9 +157,61 @@ The initial implementation includes provider-neutral orchestration helpers:
 - `apply_configuration_package(...)`
 - `export_configuration_package(...)`
 
+Portable fragments may bind deployment-specific operator input without placing
+that value in the signed reusable definition. A payload value of
+`{"$data": "requirement_key"}` references a key declared in the manifest's
+`data_requirements`. Preflight fails before invoking the owning provider when a
+reference is malformed, undeclared, or unresolved. Once supplied, Core replaces
+the reference in memory and passes only the resolved fragment to the provider.
+This mechanism is for deployment bindings and wording, not plaintext secrets:
+credential-envelope or environment references remain the normal portable
+boundary.
+
 The first concrete provider is `govoplan_access.backend.configuration_provider`.
 It supports access-owned `roles`, `groups`, and `group_role_assignments`
-fragments and applies them idempotently.
+fragments and applies them idempotently. Mail and Files also register providers
+for deployment configuration: Mail owns receipt-bound SMTP profiles and Files
+validates the deployment-owned managed-storage binding.
+
+### Deployment capability receipt
+
+The installer mounts a bounded, non-secret infrastructure receipt at the path
+named by `GOVOPLAN_DEPLOYMENT_CAPABILITIES_PATH`. Core validates that document
+once for configuration-package context and exposes typed capability and
+post-install-task records to providers. Invalid receipts fail closed. Endpoint
+metadata is sanitized, and secret fields may cross this boundary only as
+`env:VARIABLE_NAME` references.
+
+Feature providers remain responsible for their own semantics:
+
+- Mail can derive host and port from `mail.smtp`, collect missing non-secret
+  transport fields, and bind an existing credential-envelope id. It never
+  accepts or exports a username, password, token, or decrypted credential.
+- Files compares `files.storage` with the effective runtime backend, endpoint,
+  trust marker, bucket, and presence of referenced environment secrets. Storage
+  remains deployment-owned, so the provider reports `skip` when they agree and
+  blocks drift instead of rewriting process environment or storage credentials.
+- A system-scoped Mail profile requires system configuration authority. Tenant
+  scope is the conservative default.
+- Existing Mail configuration is preserved unless a reviewed fragment
+  explicitly selects `on_conflict: update`. Reapplying an unchanged fragment is
+  a no-op.
+
+Ops projects the same Core-validated receipt. It must not maintain a second
+parser with different validation or secret-handling rules.
+
+Core also defines the inverse, read-only dependency-inventory contract used
+before the installer changes one of those infrastructure capabilities. An
+enabled module registers
+`infrastructure.dependency_inventory.<module_id>` and returns bounded, stable
+references to its persisted configuration or data, a lifecycle state, scope,
+numeric metrics, and a required operator action. Providers must not return
+secrets or use this read to migrate state. The Core collector validates provider
+identity and capability coverage, orders records deterministically, and marks
+the complete inventory failed when any provider raises or violates the
+contract. Ops is the authorized projection boundary; the installer remains the
+consumer and must match installation id, freshness, completion and impacted
+capability coverage before apply.
 
 The admin wizard backend starts with these routes:
 
@@ -183,6 +235,14 @@ The admin wizard backend starts with these routes:
 10. Store import provenance, package version, supplied non-secret metadata, and
     audit events.
 
+Provider applies may commit independently. Core therefore stops at the first
+apply or health blocker and reports an explicit rollback state. A blocked
+preflight or a no-op needs no recovery; a successful multi-provider mutation
+retains the reviewed pre-apply database snapshot as its generic rollback path;
+a later-provider failure is reported as a partial apply that requires snapshot
+recovery or an explicitly supported module-owned compensation. The generic
+wizard never claims atomic cross-module undo.
+
 The wizard should display everything necessary and nothing unnecessary. Generic
 sections should cover package trust, dependency plan, required data, conflicts,
 review, and result. Module-specific fields should appear only when the selected
@@ -233,6 +293,11 @@ Exported packages should record provenance: source GovOPlaN version, module
 versions, exporter identity, timestamp, selected scope, redactions, and
 validation status.
 
+The orchestrator emits this provenance independently of provider payloads and
+lists secret requirement keys as redacted without serializing their supplied
+values. Providers still own the deeper rule that credentials, tokens, and
+decrypted envelope contents must never appear in exported fragments.
+
 ## Catalogs And Trust
 
 Configuration catalogs should follow the existing module package catalog model:

docs/CONTEXTUAL_HELP_CONTRACT.md

@@ -0,0 +1,84 @@
+# Contextual Help Contract
+
+GovOPlaN exposes context-sensitive help through `F1` and the titlebar help
+control. The shell resolves a stable help identity from the focused control,
+its containing surface, and the current route. The Docs module then projects
+the best visible user or administrator topic for that identity.
+
+## Resolution Order
+
+The WebUI resolves help in this order:
+
+1. an explicit `helpContextId` or `data-help-context-id` on the focused item
+2. the focused shared control's `interfaceId`, `helpTopicId`, and label key
+3. a containing dialog, card, administration section, or page surface
+4. the current registered route, including dynamic module routes
+5. a stable route-derived fallback when no explicit identity is available
+
+Focused field and action contexts retain the page context as
+`fallback_context`. This lets Docs show a field-specific topic when one exists
+and otherwise open the owning page or module documentation instead of a generic
+help page.
+
+## Documentation Lookup
+
+Static `DocumentationTopic` contributions announce exact contexts through
+`metadata.help_contexts`. Core publishes that catalogue with the enabled module
+manifest, allowing the shell to link directly to an exact topic when possible.
+Docs still performs the authoritative audience, permission, configured-state,
+and documentation-type filtering.
+
+Core also maps explicit route, navigation, settings, and View surface IDs to
+the module's static user or administrator documentation baseline. This makes a
+page association complete by default and gives every derived field/action
+context a useful fallback. Exact `metadata.help_contexts` remain the preferred
+authoring mechanism for consequential or unfamiliar controls.
+
+When there is no exact topic, Docs resolves the page fallback and then the first
+visible topic owned by the module. If Docs is unavailable, the shell opens the
+hosted documentation with the same context parameters.
+
+## Authoring Controls
+
+Core shared controls expose stable help metadata. Prefer these props rather
+than adding custom `F1` listeners:
+
+- `interfaceId` identifies a durable UI surface or action.
+- `helpContextId` identifies a documentation context when it differs from the
+  interface identity.
+- `helpModuleId` identifies the documentation-owning module when a shared
+  control is embedded in another module's page.
+- `helpTopicId` links directly to a module-owned documentation topic.
+- translated label keys provide deterministic field identities for ordinary
+  `FormField`, `ToggleSwitch`, search, date/time, email, button, dialog, and card
+  controls.
+- `TableActionGroup` action definitions carry the same identities so focused
+  row actions can resolve consequence-specific help.
+- `PageLayout` owns the page help scope and documentation identity for ordinary
+  headed pages. `WorkspaceLayout` owns the full-canvas workspace scope and its
+  labelled primary/content panes; pages inside it use `PageLayout` in
+  `workspace` mode and retain their own route-level help identity.
+- `PasswordField` passes its owner context and module through reveal/generate
+  actions and the shared generator dialog. Credential consumers must supply an
+  exact owner context; the generic component does not own credential policy.
+
+High-risk controls use one of the source-inventory risk classes (`authority`,
+`credential`, `disclosure`, `encryption`, `external-effect`, `irreversible`,
+`policy`, or `retention`) and require exact F1 help. The extractor infers
+obvious cases conservatively; components may declare `data-help-risk`
+explicitly or mark a reviewed ordinary control with
+`data-help-risk-reviewed="standard"`. The strict workspace gate rejects new
+unresolved high-risk debt.
+
+Module routes, public routes, settings sections, and administration sections
+may also declare `helpContextId` and `helpTopicId`. Each module must keep a
+static user/admin documentation baseline and should list its important route,
+workflow, setting, permission, and limitation identities in
+`metadata.help_contexts`.
+
+## Boundary
+
+Help identities describe presentation context; they are not authorization
+claims. Opening help never bypasses route or documentation permissions. Docs
+owns documentation projection, feature modules own their content, and Core owns
+focus capture, context resolution, and fallback routing.

docs/DATAGRID_SIZING_CONTRACT.md

@@ -1,8 +1,31 @@
 # DataGrid Sizing Contract
 
+Auto-height grids reserve no empty vertical scrollbar gutter. The table fills
+its card to the right edge; an actual constrained vertical scrollbar still
+occupies its normal space. `Card bodyLayout="table"` provides an explicit
+zero-inset surface, including with loading wrappers and padded notices. The
+Organizations/IDM browser fixtures assert row geometry, not just outer shells.
+
+The containing card removes its body padding; tables stay at `width: 100%` and
+`max-width: 100%`. Do not cancel padding with negative margins or an expanded
+`calc()` width: the table's maximum width correctly clamps that expansion, and
+the result is a visible gap. Existing cards with only a DataGrid, admin table
+surface, or connection tree inherit the same zero-inset geometry, directly or
+through a `LoadingFrame`. Its loading overlay does not count as content and
+stays within the table body. Mixed prose/form content keeps the usual insets;
+choose `bodyLayout="table"` explicitly when combining a full-width table with
+separately padded notices. Ordinary standalone tables retain their border.
+
+Deutsch: Tabellenkarten entfernen den Innenabstand am Karteninhalt. Die Tabelle
+bleibt bei 100 Prozent Breite; negative Ränder und verbreiternde `calc()`-Werte
+sind nicht erforderlich. Das gilt auch während des Ladens. Karten mit Text
+oder Formularen behalten ihren Innenabstand. Für eine randlose Tabelle neben
+einem separat gepolsterten Hinweis ist `bodyLayout="table"` ausdrücklich zu
+setzen; alleinstehende Tabellen behalten ihren Rahmen.
+
 `DataGrid` turns every declared track into a deterministic pixel layout after
 its container has a measurable width. The same contract is used on initial
-layout, container resize, persisted-layout restore, and pointer resize.
+layout, container resize, persisted-layout restore, and pointer/keyboard resize.
 
 ## Column Declarations
 
@@ -17,6 +40,77 @@ layout, container resize, persisted-layout restore, and pointer resize.
 - `maxWidth` bounds direct user growth and free/constrained compensation. In a
   cover layout it is a preferred maximum: passive tracks may exceed it when
   that is necessary to keep the table flush with its container.
+- `preferredMaxWidth` is a presentation-only automatic-fit ceiling. It retains
+  sensible initial proportions without limiting direct user resizing or
+  right-side resize compensation. A declared `maxWidth` still wins as the
+  manual limit. Cover may exceed preferred ceilings after other automatic
+  sizing targets are exhausted, preserving the no-blank-filler contract.
+- Do not give expandable text columns arbitrary `maxWidth` values merely to
+  control their initial appearance. At that limit, a resize handle can shrink
+  but cannot grow, regardless of whether the next column is resizable. Prefer
+  weighted `minmax(..., Nfr)` declarations with `preferredMaxWidth` when needed;
+  ordinary cover-mode growth may
+  create horizontal overflow without changing fixed neighbors. Campaign's
+  Recipient(s), Delivery and configurable recipient-field columns follow this
+  contract. The existing signature format is unchanged for other columns;
+  introducing a preferred ceiling invalidates only that grid's previous width
+  contract, while sort/filter preferences remain intact.
+- `columnType: "actions"` marks a custom action/control column. Canonical
+  `TableActionGroup` content is recognized automatically, even in existing
+  column declarations. Use `sticky: "end"` for the normal row-action surface.
+
+## Action Visibility and Constrained Containers
+
+Action tracks reserve the width of actual buttons, disabled-action wrappers,
+reserved empty-state slots, gaps, and cell padding. A historic `width: 72`
+preference therefore cannot clip a four-button action group. Ordinary data text
+does not participate in this content measurement; long field values do not
+silently widen all tracks. Changes to the rendered action set are remeasured.
+
+`TableActionButton` remains a compact 36 px control, including the Add action
+in an empty grid. Never stretch it with a last-column `.btn { width: 100% }`
+rule. Its shared maximum width and fixed flex basis protect against broad
+consumer button rules, which can otherwise feed stretched widths back into
+action-track measurement and consume the data area.
+
+When the full group needs more than half the scroll viewport, its measured
+minimum is capped at half the viewport and the group wraps. Explicit hard
+minima remain authoritative. Custom action groups should use wrapping-capable
+flex layouts and semantic groups, preferably composing `TableActionGroup`.
+
+The grid's physical width matches its pixel tracks, including horizontal
+overflow, so right-sticky actions remain inside the correct scroll bounds.
+If explicitly wide or persisted sticky tracks would obscure the readable data
+area, horizontal stickiness is released until space returns. No columns or
+actions are hidden: the labelled scroll region is focusable and supports native
+keyboard scrolling. Vertical header stickiness remains available.
+
+## Resizing Controls
+
+Drag a resize handle with a mouse, pen, or touch pointer. Pointer capture keeps
+the drag active when it leaves the handle. Escape or pointer cancellation
+restores the layout before that drag; releasing the pointer commits it. Losing
+window focus ends a drag without leaving the table stuck in resizing mode.
+
+Each handle is a focusable vertical separator exposing its current and allowed
+widths. Left/Right changes its width by 10 px; Shift+Left/Right uses 40 px. Enter
+or a double-click resets that column's explicit override to the declared sizing
+rules. Other columns retain their preferences, so cover/compensation constraints
+still apply. These operations only change personal browser layout, never rows.
+
+Deutsch: Spalten lassen sich mit Maus, Stift oder Touch ziehen. Escape verwirft
+den laufenden Ziehvorgang. Am fokussierten Trenner ändern Links/Rechts die Breite
+um 10 px, mit Umschalt um 40 px. Eingabe oder Doppelklick setzt die persönliche
+Breite dieser Spalte zurück. Schmale Aktionenspalten umbrechen ihre Schaltflächen;
+breite Tabellen bleiben horizontal scrollbar.
+Eine feste Maximalbreite stoppt das Vergrößern unabhängig von Nachbarspalten.
+`preferredMaxWidth` begrenzt dagegen nur die automatische Anfangsaufteilung,
+nicht persönliche Breiten oder den Ausgleich beim Ziehen. Empfänger,
+Zustellung und konfigurierte Empfängerfelder verwenden diese Darstellungsgrenze.
+Bei Bedarf verbreitert sich die Tabelle und bleibt scrollbar; persönliche
+Breiten ändern keine Kampagnendaten. Am rechten Scrollrand kann eine breite
+linke Spalte wieder verkleinert werden, weil rechte Textspalten über ihre
+bevorzugte Anfangsbreite hinaus Platz aufnehmen dürfen.
 
 ## Layout Modes
 
@@ -36,16 +130,25 @@ the column remains stopped until the pointer crosses the same boundary again.
 
 ## Persistence
 
-Only the pixel layout resulting from an explicit user resize is persisted.
+Only the pixel layout resulting from an explicit user resize is persisted,
+together with the container width at which the user selected it.
 Persisted widths are keyed by a signature containing column IDs, declared
-widths and bounds, resize affordances, sticky placement, initial fit, and resize
+widths and bounds, sort/filter/resize affordances, column type, sticky placement, initial fit, and resize
 behavior. A changed signature discards the old override and recomputes the
 declared layout.
 
 Container reconciliation is suspended while a pointer drag is active. On
 release, the already-rendered pixel layout becomes the persisted preference.
-Reconciliation may grow it to prevent underflow, but never shrinks intentional
-user overflow, so there is no drag-end snap.
+Reconciliation at that same container width never shrinks intentional user
+overflow, so there is no drag-end snap. If the surrounding layout later
+contracts, persisted tracks may shrink toward their hard minima. The layout
+retains only the amount of horizontal overflow deliberately created by the
+user; an exact-cover layout therefore remains exact-cover at narrower widths.
+Legacy snapshots from the former hard-pixel persistence contract are discarded
+once and recomputed from the declared column layout. The current `v3` signature
+also discards old snapshots that predate action and header-control minima;
+sort/filter preferences remain intact. Measured action widths are not included
+in the signature, so changing rows does not erase user sizing intent.
 
 ## Regression Matrix
 
@@ -56,6 +159,7 @@ user overflow, so there is no drag-end snap.
 - hard-minimum horizontal overflow;
 - fixed-only cover grids;
 - persisted overrides under growth and viewport pressure;
+- responsive contraction of persisted layouts without losing deliberate overflow;
 - stale layout signatures;
 - first and middle-column right-side compensation;
 - last-resizable-column overflow, underflow stop, and reverse-pointer boundary;
@@ -65,3 +169,11 @@ user overflow, so there is no drag-end snap.
 
 `webui/tests/data-grid-actions.test.tsx` also verifies the rendered fixed-cover
 shape and guards against reintroducing a synthetic buffer cell.
+
+`webui/conformance/tests/data-grid-layout.spec.ts` exercises the real rendered
+grid with deliberately undersized action preferences, constrained containers,
+horizontal scrolling, changing/empty action sets, keyboard and pointer resizing,
+Escape cancellation, remount persistence, responsive contraction and restoration,
+free/content mode, and constrained compensation. Run with
+`npm run test:conformance -- data-grid-layout.spec.ts`; its isolated test server
+is stopped automatically afterwards.

docs/DATA_SUBJECT_REQUESTS.md

@@ -0,0 +1,91 @@
+# Data-Subject Request Contract
+
+This document defines the provider-neutral workflow for access and erasure
+requests. It is an operational control and evidence mechanism. It does not
+replace legal review, identity verification, retention policy, or the
+institution's statutory response process.
+
+## Ownership
+
+Core owns the request aggregate, lifecycle API, optimistic concurrency,
+provider discovery, export manifest, execution orchestration, and audit event
+names. Modules that store subject-related data own their search, explanation,
+retention, and mutation behavior through a `privacy.dsar.<module>` capability.
+Core never scans module tables or guesses how a foreign resource may be
+erased.
+
+Access owns the first provider. It finds tenant memberships plus safe account,
+identity, assignment, API-key, and session metadata. It does not export secret
+hashes, session tokens, IP addresses, or browser fingerprints. Tenant-local
+membership data can be anonymized and authentication material can be revoked.
+Global accounts and identities require manual system-level review because they
+may serve more than one tenant.
+
+## Lifecycle
+
+1. A privacy officer records a verified selector, purpose, legal basis, due
+   date, and internal reference.
+2. Search invokes every available tenant capability independently. A provider
+   failure is isolated and recorded; it cannot turn an incomplete search into
+   a successful one.
+3. The JSON export contains the request, records, provider runs, coverage,
+   retention reasons, execution evidence, and a SHA-256 manifest digest.
+4. An erasure request produces stable provider-owned actions. Immutable
+   evidence generates an explicit non-executable `retain` decision.
+5. Execution accepts only selected executable actions from the current plan.
+   It requires `If-Match`, the current resource revision, the dedicated erase
+   permission, and the exact `ERASE <request-id>` confirmation phrase.
+6. Provider execution is idempotent. Completed or unchanged effects remain
+   durable in the request's execution evidence.
+
+The API is rooted at
+`/api/v1/admin/privacy/data-subject-requests`. Access exposes the independent
+permissions `access:privacy:read`, `access:privacy:manage`,
+`access:privacy:export`, and `access:privacy:erase`; the built-in privacy
+officer role contains all four.
+
+## Provider Rules
+
+A provider must:
+
+- enforce tenant ownership for every record and action;
+- return stable, unique resource and action identities;
+- avoid credentials, hashes, tokens, unnecessary telemetry, and unrelated
+  third-party data;
+- distinguish mutable personal data from immutable institutional evidence;
+- state a retention reason for immutable evidence;
+- propose manual review instead of an automatic action when authority is
+  ambiguous or a resource spans tenants;
+- return exactly one execution result per requested action;
+- make execution idempotent and avoid committing the caller's transaction;
+- keep all actual mutations inside the owning module.
+
+Each active module without a DSAR provider is listed in coverage. This is a
+deliberate fail-visible state, not proof that the module stores personal data.
+An institution may call an export complete only after it has reviewed both the
+provider runs and that coverage list.
+
+## Retention And Evidence
+
+Erasure and retention are separate decisions. Stable object IDs, authorization
+history, function incumbency, formal decisions, delivery evidence, and audit
+records may remain necessary for accountability. Providers expose those items
+with a concrete reason and Core prevents them from being selected as executable
+actions. Policy may further restrict an action, but it must never silently
+loosen a provider's retention decision.
+
+All lifecycle mutations and exports produce tenant audit events. The request
+stores an evidence digest after every revision. This digest detects accidental
+or unauthorized mutation of the aggregate; it is not a digital signature or a
+substitute for signed recovery evidence.
+
+## Current Limits
+
+- Access is the first native provider. Other enabled modules appear in the
+  coverage list until they add a provider or an explicit no-subject-data
+  declaration is standardized.
+- Verification of the requester's identity and statutory deadline escalation
+  remain institutional workflows outside this API.
+- Global account or identity erasure is deliberately manual.
+- Exports are JSON. A human-readable signed response package remains a later
+  Reporting/Templates integration.

docs/DEPLOYMENT_OPERATOR_GUIDE.md

@@ -7,6 +7,16 @@ files.
 
 ## Runtime Configuration Contract
 
+Worker and queue observability is provider-neutral. Runtime modules register a
+bounded `RuntimeWorkStatusProviderRegistration` with Core; the Ops module
+projects its sanitized status without importing Celery, Redis, or module job
+implementations. Providers must use explicit `null` values for unsupported
+queue depth, active/reserved work, failure count, and heartbeat evidence. An
+unavailable metric must never be interpreted as zero or as proof of health.
+The standard Core adapter reports the configured Celery/Redis runtime and
+combines its bounded inspection result with registered worker heartbeat and
+stale-threshold evidence.
+
 Self-hosted installability follows the staged approach documented in
 `SELF_HOSTED_INSTALLABILITY.md`: generate an explicit env template, validate it,
 run production-like rehearsal with Compose-backed dependencies, then use the
@@ -64,6 +74,19 @@ Operator rule: take a database backup before applying migrations or destructive
 module retirement. For non-SQLite databases, configure deployment-specific
 backup/restore hooks for the module installer.
 
+#### Ownership-history upgrade repair
+
+Core revision `c58a2d7e9f10` repairs existing ownership-transfer tables that
+predate the `decisions` column. Such installations can otherwise return HTTP
+500 from `/api/v1/ownership/transfers`, including Campaign Settings. Apply the
+normal forward migrations after taking a backup; do not stamp a revision or
+recreate the table. The repair is available on both migration tracks, adds only
+the missing non-null JSON column, and initializes old rows with an empty list.
+It preserves owners, approvals, transfer states, revisions, timestamps, and any
+existing decision history. Historical decisions are not reconstructed or
+invented. Downgrading this repair retains the additive column and its evidence.
+Verify that ownership-transfer listing and Campaign Settings load after upgrade.
+
 ### PostgreSQL Production Target
 
 PostgreSQL is the primary development and production target. SQLite remains
@@ -278,12 +301,15 @@ through the same trusted address range.
 | --- | --- |
 | `GOVOPLAN_MODULE_PACKAGE_CATALOG_URL` or `GOVOPLAN_MODULE_PACKAGE_CATALOG` | Module package catalog source. |
 | `GOVOPLAN_MODULE_PACKAGE_CATALOG_TRUSTED_KEYS_FILE` | Preferred production keyring path. |
-| `GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL` | Approved catalog channel, for example `stable`. |
+| `GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS` | Comma-separated approved catalog channels, for example `stable`. The legacy singular name remains readable during migration. |
 | `GOVOPLAN_LICENSE_TRUSTED_KEYS_FILE` | Trusted license issuer keyring path. |
 | `GOVOPLAN_LICENSE_ENFORCEMENT` | Enables license enforcement when set to `true`. |
 
 Trust roots are deployment-managed and should not be editable through the
-running WebUI.
+running WebUI. When no catalog override is configured, the Admin package
+directory uses GovOPlaN's public stable catalog and the trust anchor bundled
+with the installed Core release. Production operators may still pin a newer or
+institution-specific catalog/keyring explicitly with the settings above.
 
 ### Mail Test Credentials
 
@@ -395,6 +421,27 @@ To stop PostgreSQL and Redis when the launcher exits:
 GOVOPLAN_STOP_PROFILE_DEPENDENCIES_ON_EXIT=1 tools/launch/launch-production-like-dev.sh
 ```
 
+## Development WebUI Dependency Caches
+
+The application and browser-conformance harness share installed JavaScript
+packages but must not share Vite's optimized-dependency cache. The application
+uses `webui/node_modules/.vite/govoplan-app`; the conformance harness uses
+`webui/node_modules/.vite/govoplan-conformance`. Keep these explicit sibling
+directories when adding development or test configurations. Setting a different
+Vite `root` alone does not isolate this cache.
+
+A shared cache can make otherwise healthy Workflow, Dataflow or deferred editors
+show “The resource could not be loaded.” The browser then reports an asset such
+as `@xyflow_react.js` with HTTP 504 `Outdated Optimize Dep`, while the corresponding
+API still returns HTTP 200. This is not a missing workflow permission or a reason
+to rerun a pipeline. Preserve unsaved work, let the existing development server
+reload the corrected configuration (or restart that WebUI server), then reload
+the browser. Do not clear application data, change grants or restart delivery
+workers to repair a frontend dependency cache.
+
+Run `npm run test:vite-cache-isolation` in `govoplan-core/webui` to verify the real
+resolved Vite configurations without starting servers or overwriting caches.
+
 ## Module Install/Uninstall Operations
 
 Use Admin > System > Modules for planning. The running API server validates and

docs/DOCUMENTATION_MAP.md

@@ -17,6 +17,14 @@ operator, and roadmap pages.
 | Action/effect automation layer | `ACTION_EFFECT_AUTOMATION_LAYER.md` | Action/effect contracts, consequence preview, runner semantics, and module boundary for automation. |
 | External references and integration maturity | `EXTERNAL_REFERENCES_AND_INTEGRATION_MATURITY.md` | Stable external identity and cumulative connector maturity; configured source authority is defined by the meta target architecture. |
 | Institutional context and governed references | `INSTITUTIONAL_CONTEXT_CONTRACT.md` | Shared temporal, actor/representation, institution, mandate, service, party, decision, evidence, legal-basis, information-governance, presentation, and geo DTO/provider contracts. |
+| Provider-neutral record filing | `RECORDS_FILING_CONTRACT.md` | Exact source-revision identity, current source authorization, idempotent filing, capability discovery, and ownership boundary. |
+| Ticket routing and Case escalation | `TICKET_INTEGRATION_CONTRACTS.md` | Optional fail-open routing, replay-safe Case handoff, authorization, evidence, and ownership boundaries. |
+| Temporal data read context | `TEMPORAL_DATA_CONTEXT.md` | Valid-time and recorded-time titlebar selection, HTTP/cache contract, security boundary, and module-adoption rule. |
+| Cross-module information governance adoption | `INFORMATION_GOVERNANCE_ADOPTION.md` | Manifest evidence and enforcement rules for temporal browsing, purpose-aware access, retention, and institutional context. |
+| Data-subject access and erasure requests | `DATA_SUBJECT_REQUESTS.md` | Provider-owned search and mutation, explicit coverage, governed export, retained evidence, permissions, and idempotent execution. |
+| Context-sensitive F1 help | `CONTEXTUAL_HELP_CONTRACT.md` | Focus, route, module-manifest documentation contexts, Docs projection, and hosted fallback. |
+| Semantic documentation subjects | `SEMANTIC_DOCUMENTATION_SUBJECTS.md` | Stable configured-artifact identity, safe provider discovery, revision review, authorization, and lifecycle semantics. |
+| German localization and help quality gate | `LOCALIZATION_AND_HELP_QUALITY.md` | German reference locale, new-installation default, catalog completeness, automatic page associations, and explicit-help review priorities. |
 | Postbox E2EE target architecture | `POSTBOX_E2EE_ARCHITECTURE.md` | Strategic encrypted postbox/mailbox model, key ownership, role mailbox semantics, and retraction limits. |
 | Shared state, runtime coordination, and recovery | `STATE_AND_RECOVERY_CONTRACT.md` | State profiles, object storage, node registration/drain, fenced leases, migration ordering, and recovery evidence. |
 | Module lifecycle recovery | `MODULE_LIFECYCLE_RECOVERY.md` | Installer/live-graph recovery modes, deployment fence, evidence, retry blocking, and operator reconciliation. |
@@ -37,7 +45,9 @@ operator, and roadmap pages.
 | Topic | Canonical document | Notes |
 | --- | --- | --- |
 | Product roadmap and module routing | `GOVOPLAN_MASTER_ROADMAP.md` | Product-level sequencing, implementation gates, issue routing, and missing-module decisions. |
-| Institutional governance target | `govoplan/docs/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md` | Cross-product semantic layers, source-authority modes, candidate Mandates/Services/Parties/Decisions boundaries, and migration sequence. |
+| Stable platform ideas | `govoplan/docs/strategy/PLATFORM_CORE_IDEAS.md` | Cross-product thesis, canonical distinctions, product experience rule, maturity rule, and decision test. |
+| Current cross-product reconciliation | `govoplan/docs/strategy/STRATEGY_STATUS.md` | The only current prose status source; generated evidence and Gitea remain authoritative inputs. |
+| Institutional governance target | `govoplan/docs/architecture/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md` | Cross-product semantic layers, source-authority modes, candidate Mandates/Services/Parties/Decisions boundaries, and migration sequence. |
 | UI/UX decisions | `UI_UX_DECISION_LEDGER.md` | Binding guided-UI decisions, open decisions, impact index, and review checklist. |
 | Core interface migration | `INTERFACE_PATTERN_MIGRATION.md` | Core-owned settings, credential, retention, lifecycle, and shared-component evidence for the product pattern language. |
 | Interface ethics and design doctrine | `INTERFACE_ETHICS_AND_DESIGN_DOCTRINE.md` | Product-level doctrine for context, decision, consequence, contestability, responsibility, and traceability. |

docs/EXTERNAL_REFERENCES_AND_INTEGRATION_MATURITY.md

@@ -40,7 +40,7 @@ must not imply that GovOPlaN holds an authoritative copy.
 Integration maturity states what an adapter is capable of doing. It does not
 decide which system owns truth for a configured object or field group. A
 binding separately selects one of the source-authority modes defined by the
-[institutional governance target architecture](../../govoplan/docs/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md):
+[institutional governance target architecture](../../govoplan/docs/architecture/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md):
 
 - `native_authoritative`
 - `external_authoritative`

docs/GITEA_ISSUES.md

@@ -13,4 +13,4 @@ tools/gitea/gitea-sync-wiki.py --help
 
 Canonical documentation:
 
-- `/mnt/DATA/git/govoplan/docs/GITEA_ISSUES.md`
+- `/mnt/DATA/git/govoplan/docs/project/GITEA_ISSUES.md`

docs/GOVERNANCE_MODEL.md

@@ -221,7 +221,8 @@ Admin lists use bounded container grids:
 - recipient import with column mapping;
 - session/device revocation UI;
 - backup/restore, monitoring, and update procedures;
-- DSAR workflows and evidence bundle verifier;
+- additional module providers and signed human-readable response packages for
+  the implemented DSAR workflow described in `DATA_SUBJECT_REQUESTS.md`;
 - campaign ownership transfer workflow;
 - policy impact analysis before delete/disable/unshare/change;
 - LDAP/OIDC/SAML provisioning;

docs/GOVOPLAN_MASTER_ROADMAP.md

@@ -10,15 +10,15 @@ gates. Issues are the active backlog; this document is durable architecture
 planning context and should be mirrored to the Gitea wiki.
 
 The meta repository's
-[Connected Governance Platform Roadmap](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/CONNECTED_GOVERNANCE_PLATFORM_ROADMAP.md)
+[GovOPlaN Roadmap](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/strategy/ROADMAP.md)
 describes the corresponding cross-product stakeholder visions, configurable
 service and operating configurations, connected outcome stories, and
 capability horizons. The selected five-stage delivery sequence and its gates
 are in the meta repository's
-[Reference Journey Program](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/REFERENCE_JOURNEY_PROGRAM.md).
+[Reference Journey Program](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/strategy/REFERENCE_JOURNEY_PROGRAM.md).
 The semantic target, source-authority modes, and reconciliation with the
 implemented platform are in the meta repository's
-[Institutional Governance Target Architecture](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md).
+[Institutional Governance Target Architecture](https://git.add-ideas.de/GovOPlaN/govoplan/src/branch/main/docs/architecture/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md).
 Those product documents are canonical; this Core roadmap remains their
 technical sequencing and module-routing companion.
 

docs/INFORMATION_GOVERNANCE_ADOPTION.md

@@ -0,0 +1,131 @@
+# Information Governance Adoption
+
+## Platform Rule
+
+Temporal browsing, purpose-aware access, retention, and institutional context
+are platform-wide information-governance dimensions. Every module receives the
+same contract by default. A module may claim `partial` or `enforced` only with
+repository-owned object scope, evidence, and limitations; it may claim
+`not_applicable` only when the dimension genuinely does not apply.
+
+Historical business data is always authorized under the current security
+state. No module may use a historical permission, membership, role, function
+assignment, or policy projection to weaken present-day access.
+
+Platform-wide adoption is tracked in
+[GovOPlaN #40](https://git.add-ideas.de/GovOPlaN/govoplan/issues/40), with
+temporal reads detailed in
+[GovOPlaN #39](https://git.add-ideas.de/GovOPlaN/govoplan/issues/39).
+
+## Manifest Declaration
+
+`ModuleManifest.information_governance` publishes four dimensions:
+
+- `temporal_browsing`;
+- `purpose_aware_access`;
+- `retention`;
+- `institutional_context`.
+
+Each dimension declares:
+
+- adoption: `not_applicable`, `contract_only`, `partial`, or `enforced`;
+- object types covered;
+- repository-local test/documentation evidence;
+- the remaining limitation for `contract_only` or `partial`.
+
+The default is intentionally `contract_only`. It applies the platform rule
+without pretending that existing domain queries and effects already enforce
+it. `reference_ready`, `supported`, and `lts` modules cannot retain an
+applicable dimension below `enforced`.
+
+## Read Contract
+
+For every persistent domain object, the owner classifies the read:
+
+1. **Current-only:** historical semantics do not exist and the API says so.
+2. **Valid-time:** select facts effective now or at the requested instant.
+3. **Bitemporal:** additionally select only revisions known by `recorded_at`.
+4. **All-validity:** return effective revisions in a bounded history view.
+
+The Core temporal middleware supplies the request context. Owners apply it in
+repositories or query helpers, include it in cache keys, return evaluated
+context, and test current/at/all plus recorded-time boundaries. Search,
+reporting, exports, selectors, counts, and drill-through must use the same
+projection as the owning list/detail API.
+
+## Purpose-Aware Access Contract
+
+Permission establishes a technical action ceiling. Purpose-aware access asks
+whether this actor, represented capacity, case/work item, legal basis, and
+declared use may access this object now.
+
+- A client-supplied purpose is an assertion, never authority by itself.
+- The owner or Policy capability validates the purpose and returns explainable
+  provenance.
+- Sensitive access can require case assignment, mandate, reason capture,
+  approval, or break-glass evidence.
+- Search, selectors, reporting, exports, background jobs, and connectors apply
+  the same decision.
+- Audit records the validated purpose identifier and decision reference, not
+  unnecessary content.
+
+## Retention Contract
+
+Every persistent object declares an owner, retention class or policy reference,
+trigger, start instant, hold behavior, review/disposition action, and evidence.
+Retention is not a generic timestamp deletion job.
+
+- Domain owners enumerate and execute their own effects through a typed
+  retention provider.
+- Policy resolves inherited ceilings and simulation.
+- Records owns record disposition; Files owns byte/object effects; Audit owns
+  audit-detail behavior; external providers declare their own effect and
+  recovery semantics.
+- Dry-run, legal hold, exact revision, idempotency, outcome unknown,
+  reconciliation, correction, and destruction evidence are mandatory for
+  consequential removal.
+
+## Institutional Context Contract
+
+Consequential objects and effects carry the relevant tenant, institution,
+organization unit, function, mandate/jurisdiction, service/case/work item,
+party/representation, decision, and record references. Context is minimized to
+what the operation needs. Organizational membership is not itself permission
+or mandate.
+
+Events, automation intents, audit evidence, records, and external effects retain
+the same governed context envelope or an exact reference to it. Consumers must
+not reconstruct authority later from mutable current structures.
+
+## Adoption Order
+
+1. Inventory every domain list/detail/search/export/effect and classify all
+   four dimensions.
+2. Migrate institutional owners first: Access, IDM, Organizations, Mandates,
+   Services, Parties, Cases, Approvals, Committee, Decisions, Voting, and
+   Records.
+3. Migrate communication and content: Addresses, Distribution Lists, Campaign,
+   Postbox, Mail, Calendar, Files, Templates, and Forms Runtime.
+4. Migrate data projections: Connectors, Datasources, Dataflow, Reporting,
+   Search, Risk Compliance, and Dashboard.
+5. Migrate workflow/task/background/provider operations and prove that no
+   asynchronous path drops context.
+6. Advance manifest claims only after owner tests and browser/reference-journey
+   evidence pass.
+
+The generated platform inventory reports adoption counts and module details.
+Gitea tracks individual migrations; the declaration is evidence and a maturity
+gate, not a substitute for implementation.
+
+## Definition Of Enforced
+
+A dimension is `enforced` only when:
+
+- all declared object types and public reads/effects use it;
+- list/detail/count/search/export/worker behavior is consistent;
+- cache and pagination semantics cannot cross contexts;
+- absence, invalid values, and inaccessible referenced context fail safely;
+- tests cover current, historical, unauthorized, replay, and module-absence
+  combinations appropriate to the dimension;
+- user/admin documentation explains behavior and limitations;
+- the manifest cites those tests and docs.

docs/INSTITUTIONAL_CONTEXT_CONTRACT.md

@@ -8,6 +8,9 @@ The shared contract lives in `govoplan_core.core.institutional`.
 Core owns reference shapes and provider protocols only. It does not own shared
 Mandate, Service, Party, Decision, evidence, or geography tables. Domain modules
 own persistence and authorization; optional capabilities resolve the references.
+Interactive reads use the separate platform temporal-data context documented in
+`TEMPORAL_DATA_CONTEXT.md`; it never changes current authorization or supplies
+mutation dates.
 
 ## Envelope
 

docs/INTEGRITY_PERFORMANCE_CONTRACT.md

@@ -0,0 +1,101 @@
+# Integrity-preserving performance contracts
+
+These are implementation guarantees and regression-test boundaries, not a
+security certification or production load-test result. Feature-specific policies
+and help remain in the owning modules' English and German documentation topics.
+
+## Refreshes, edits and table rendering
+
+The shell runs at most one module-load refresh per authority generation. Multiple
+invalidations coalesce into one trailing refresh; stale results and errors cannot
+replace a newer generation. Focus refreshes are throttled to five seconds and
+focus/visibility refreshes on hidden pages are suppressed; explicit module
+invalidations still trigger a read. Authentication/tenant reset disposes the
+previous controller. Authentication and authorization checks are not cached away.
+
+Editable modules must reconcile a save against the submitted draft and accepted
+server revision: edits made while a request is pending remain dirty. Completion
+must be fenced by security-relevant authority and selection, not merely an auth
+object's reference identity. A harmless profile refresh must not discard an
+accepted newly created ID and invite a duplicate create. An old account's mutation
+continuation must not reload its catalogue into the current account's page.
+
+DataGrid precomputes first-occurrence row indices once for client sorting and
+filtering. Duplicate object references, primitive values, `NaN` and sparse arrays
+retain `Array.indexOf` behavior. Comparator order, visible pagination, server-side
+pagination, sizing and resize rules are unchanged. Do not replace this with a
+last-occurrence map or change ordering as a side effect of an optimization.
+
+## Conditional responses
+
+The shared JSON GET middleware performs route handling, including authorization,
+before considering `If-None-Match`. It buffers only responses up to 1 MiB for a
+body-derived ETag. Known larger responses bypass buffering; an unknown-length
+stream crossing that limit replays its exact prefix and streams the remainder.
+No data is truncated and no large joined copy is created. The crossing chunk is
+already producer-owned: this is not a process-wide or route-output memory limit.
+Empty chunks do not accumulate. Large responses may no longer receive a
+middleware-generated ETag/304; explicit route ETags remain intact. Small-response
+cache semantics and credential/language/context `Vary` fields are retained.
+
+## Shared helpers and concurrency
+
+Central helpers replace exact live-code duplicates only. Actor precedence,
+whitespace handling and service-account differences remain explicit owner choices;
+historical migration code is not redirected to mutable runtime helpers. Connector
+search ACL token projection keeps its first-seen ordering and existing 500-token
+cap, stopping work once that cap is reached. Provider schemas are inferred in one
+pass without storing a second list of every column value.
+
+The keyed-list three-way merge retains insertion anchors when unrelated fields
+change. Concurrent additions use deterministic ordering; contradictory anchors
+produce a collection-order conflict instead of silently relocating an item.
+No existing endpoint is newly opted into merge behavior by this change.
+
+SQL JSON authorization predicates support the explicitly tested SQLite and
+PostgreSQL dialects, retain exact string membership and reject unsupported
+dialects. Apply tenant and authorization predicates before counting/pagination;
+never page a broader result first and filter away unauthorized records afterward.
+SQLite execution and PostgreSQL SQL compilation are not substitutes for a
+deployment's PostgreSQL concurrency and representative-data load tests.
+
+## Migration connection URLs
+
+Alembic preserves the configured database URL exactly in online and offline
+migration modes, including percent-encoded credentials and PostgreSQL Unix-socket
+paths. Escaping applies only at its ConfigParser boundary; operators must not
+double-escape `%` in `DATABASE_URL` or alter working connection credentials to
+work around interpolation errors. This does not change the target database,
+authentication, TLS policy, or migration contents.
+
+## Deutsch: Integrität vor Geschwindigkeit
+
+Der zentrale Modul-Refresh bündelt gleichzeitige Auslöser und verwirft veraltete
+Ergebnisse einschließlich Fehlermeldungen. Ein Wechsel von Anmeldung oder Mandant
+beendet die bisherige Generation. Fokusaktualisierungen sind auf einen Auslöser
+je fünf Sekunden begrenzt. Berechtigungsprüfungen bleiben erhalten.
+
+Speicherantworten dürfen zwischenzeitliche Bearbeitungen nicht überschreiben.
+Ein unveränderter Berechtigungskontext mit einem neuen Profilobjekt darf eine
+bereits bestätigte neue ID oder Revision nicht verwerfen. Umgekehrt dürfen alte
+Anfragen nach einem Kontowechsel keine Daten in den neuen Kontext übernehmen.
+Die DataGrid-Optimierung erhält Reihenfolge, Filter-, Seiten- und Größenverhalten.
+
+Die ETag-Middleware puffert höchstens 1 MiB Nutzdaten zuzüglich eines bereits vom
+Erzeuger gelieferten Grenz-Chunks. Größere Antworten werden vollständig weitergereicht,
+nicht abgeschnitten; automatisch erzeugte ETags können dabei entfallen.
+Autorisierung läuft auch bei bedingten Anfragen. Das ist keine allgemeine
+Speicherbegrenzung für Routen oder Prozesse.
+
+Gemeinsame Helfer erhalten die bisherigen fachlichen Unterschiede. Listen-Merges
+bewahren Einfügepositionen oder melden widersprüchliche Reihenfolgen explizit als
+Konflikt. Datenbankseitige Autorisierung erfolgt vor Zählung und Seitenauswahl.
+Regressionstests belegen diese Verträge; reale Provider-, PostgreSQL- und Lasttests
+in einer repräsentativen Umgebung bleiben Teil der Betriebsfreigabe.
+
+Alembic übernimmt die konfigurierte Datenbank-URL in Online- und Offline-Läufen
+unverändert, einschließlich prozentkodierter Zugangsdaten und PostgreSQL-
+Unix-Socket-Pfade. Die Maskierung erfolgt ausschließlich an der ConfigParser-
+Grenze; `%` in `DATABASE_URL` nicht doppelt maskieren und funktionierende
+Zugangsdaten nicht als Umgehung ändern. Zieldatenbank, Anmeldung, TLS-Vorgaben und
+Migrationsinhalte bleiben unverändert.

docs/INTERFACE_PATTERN_MIGRATION.md

@@ -13,10 +13,60 @@ domain modules own their compositions.
 | Reusable credentials | Repeated administration with an adaptive create/edit dialog, optional password generator, and destructive confirmation | Secret values are write-only; generated candidates use the browser cryptographic API without a weak fallback and do not replace the field until explicitly confirmed; scope/permission blockers name the required action, responsible actor, and destination; unavailable row actions remain keyboard-explainable | `CredentialEnvelopeManager.tsx`, shared `PasswordField`, `PasswordGeneratorDialog`, `ActionBlockerHint`, `Button`, `TableActionGroup`, and `ConfirmDialog` |
 | Retention policy | Effective-policy editor with inherited source paths and typed, narrowing-only controls | Parent locks and missing write authority are explicit; the save action distinguishes locks, missing target, loading, clean draft, and active save | `RetentionPolicyManagement.tsx`, policy logic tests, `test-core-interface-patterns.mjs` |
 | Module lifecycle | Guided operator projection over durable installer-queue evidence | Preflight, handoff, progress, stale evidence, recovery, and rollback consequences remain visible | Admin module lifecycle tests and the Core installer-queue contract |
-| Shared configuration primitives | Cross-module component contract | Dialog focus, blocker structure, disabled-action focus, contextual help, unsaved changes, confirmation, loading, alerts, problem lists, and policy provenance are centralized | Core component tests and module-permutation build |
+| Shared page frame | Domain-neutral headed page layout used by Core and optional modules | Standalone, workspace, and embedded modes make inset and scroll ownership explicit; sticky heading, rich descriptions, route actions, notices, loading, narrow-layout collapse, and contextual-help identity are centralized; composite administration workspaces may delegate the visible heading to their contributed panel while retaining the same frame; `AdminPageLayout` composes the contract | `PageLayout.tsx`, `page-layout.test.tsx`, Core Settings, Access administration, Docs, Mail bounce processing, Dashboard, Ops, Campaign, and `check-shared-webui-layouts.py` |
+| Full-canvas workspace | Navigation/content and list/detail canvases that own pane geometry and scrolling | Navigation and split variants, primary-pane width, pane-owned or contained scrolling, responsive stacking or navigation collapse, pane labels, and contextual-help identity are centralized without encoding domain navigation | `WorkspaceLayout.tsx`, `workspace-layout.test.tsx`, Core Settings, Access administration, Docs, Organizations, Campaign, Templates, Approvals, and `check-shared-webui-layouts.py`; the raw-workspace exception baseline is empty |
+| Full-height module frame | Outer module landmark and viewport/container sizing | `WorkspaceFrame` centralizes surface, overflow, box sizing, accessible naming, help identity, and application-viewport height so modules do not copy the `100vh - shell` frame | `WorkspaceFrame.tsx`, `layout-primitives.test.tsx`, Dataflow, Workflow, Datasources, Distribution Lists, Notifications, Tasks, Scheduling, Forms, Portal, Projects, Records, and Reporting |
+| Responsive action toolbar | Domain-neutral action and filter grouping for pages, workspaces, editors, and overlays | Density, surface, grouping, flexible space, accessible naming, toolbar help identity, and responsive wrapping are centralized while modules retain action wording, authority, and consequence | `ActionToolbar.tsx`, `layout-primitives.test.tsx`, WYSIWYG, Calendar, Files, Forms, Templates, and the product-wide primitive check |
+| Semantic page and pane action bars | Overview, collection, detail, editor, and workspace intent declared independently from frame geometry; full-canvas panes add workspace/collection/detail/editor scope | Core renders guarded Reload in the right-aligned group immediately before Create/primary actions; editor persistence owns clean, dirty, invalid, saving, failed, and conflict feedback plus guarded Discard and far-right Save; destructive actions occupy an explicit named boundary; read-only surfaces do not invent Save | `PageActionBar.tsx`, `WorkspaceActionBar.tsx`, `PAGE_LAYOUT_USAGE_GUIDELINES.md`, component and browser conformance, every headed page and full-canvas workspace, and the discovery-based `check-shared-webui-layouts.py` |
+| Catalogue and state composition | Search/filter bars, selectable navigation lists, count badges, and empty/blocked/error panels | Width, surface, wrap, selection geometry, title/description truncation, numeric emphasis, state sizing, tone and action placement are centralized; modules retain query behavior, object state and consequences | `FilterBar.tsx`, `SelectionList.tsx`, `CountBadge.tsx`, `StatePanel.tsx`, `layout-primitives.test.tsx`, and list/detail modules across Cases, Committee, Dataflow, Forms, Notifications, Portal, Postbox, Projects, Records, Reporting, Tasks, Templates, and Workflow |
+| Content and form grids | Equal-column content, field, and native-form geometry | Explicit 1–4 columns, standard gaps, item spans, alignment, and named narrow/workspace/standard/wide collapse points replace generic and module-prefixed copies; unequal domain tracks remain local | `ContentGrid.tsx`, `layout-primitives.test.tsx`, Core dashboard/settings/mail, Calendar dialogs, Forms editor, Datasources, Postbox, Campaign, administration surfaces, and the product-wide primitive check |
+| Content sections | Repeated editor/detail section surfaces | Border, surface, compact/default density, stacked flow and block rhythm are centralized without encoding section contents | `ContentSection.tsx`, `layout-primitives.test.tsx`, Datasources, Distribution Lists, Templates, Dataflow, and Workflow |
+| Form sections | Reusable heading/description/action/content grouping inside forms | Plain, separated, and panel variants centralize hierarchy and narrow action placement without moving validation, permissions, values, or domain wording into Core | `FormSection.tsx`, `layout-primitives.test.tsx`, Addresses contact editing, and Quick Access preferences |
+| Metric groups and drill-downs | Reusable responsive grouping around metric cards with an explicit optional detail affordance | Fixed one-to-five and auto-fit columns, minimum card widths, density, block/inset/zero spacing, and named collapse points replace the product-wide `metric-grid` class and cross-module dashboard overrides; typed link or in-page drill-downs name their destination while summary-only, non-enumerable, derived, or privacy-suppressed values remain inert | `MetricGrid.tsx`, `MetricCard.tsx`, `metric-card.test.tsx`, `layout-primitives.test.tsx`, Core and Dashboard summaries, administration, Campaign, Ops, Files, Search, and dashboard widgets |
+| Description lists | Semantic property and fact presentation | Stacked and inline variants, one-to-five list columns, density, term width, wrapping, and responsive collapse replace both `admin-details-grid` and `detail-list`; `DescriptionItem` preserves native `dt`/`dd` anatomy | `DescriptionList.tsx`, `layout-primitives.test.tsx`, Access and Tenancy administration, Audit, Policy, Campaign reports/imports, Docs, Settings, Ops, and Reporting |
+| Dialog anatomy | Shared outer dialog plus composable body and footer regions | Size and administration variants, body padding, descriptions, notices, fixed action wrapping, native form flow, and section grouping are centralized; focus trapping and stack lifecycle remain unchanged | `Dialog.tsx`, `DialogAnatomy.tsx`, `dialog-focus.test.tsx`, `layout-primitives.test.tsx`, Addresses, Calendar, Records, Datasources, Distribution Lists, Files, and Templates |
+| Definition-editor visuals | Reusable graph palette, canvas chrome, node icon/port geometry, empty overlay and floating activity state | Core owns visual and responsive anatomy while node/edge types, validation, execution, provenance and workflow semantics remain in Dataflow or Workflow | `DefinitionPalette.tsx`, `DefinitionNodeIcon.tsx`, `FloatingStatus.tsx`, shared definition styles, Dataflow and Workflow structure/build checks |
+| Shared configuration primitives | Cross-module component contract | Dialog focus, blocker structure, disabled-action focus, route/page/field/action F1 help, unsaved changes, confirmation, loading, alerts, problem lists, and policy provenance are centralized | Core component tests, `CONTEXTUAL_HELP_CONTRACT.md`, and module-permutation build |
+| Measured operation feedback | `LoadingFrame` over existing content | Use `indicator="none"` with native measured progress for long-running operations; `progress={null}` means unknown, never a synthetic percentage. Keep dialog content inert and close controls disabled until success or error. Existing consumers retain their loading indicator. | Files archive inspection/extraction, layout primitive tests, managed-archive browser conformance |
 
 ## Boundary
 
+Side-rail customization uses the shared `NavigationPreferenceEditor` for system,
+tenant, personal, and View layouts. Modules and labelled separators share one
+ordered list, with pointer drag-and-drop, keyboard reordering, and explicit
+add/remove actions. Consumers retain persistence and dirty-state ownership;
+mounting the editor does not create a draft change. See
+`NAVIGATION_LAYOUT_CONTRACT.md` for inheritance, locked items, optional-module
+preservation, and collapsed-rail grouping.
+
+Action columns use `TableActionGroup` or declare `columnType: "actions"` when
+their composition differs. DataGrid owns measured action minima, initial column
+allocation, persistent resizing, and local horizontal scrolling; consumers must
+not compensate with clipped overflow or copied fixed widths. See
+`DATAGRID_SIZING_CONTRACT.md`. Dialog forms use `DialogForm` and `FormGrid` inside
+the shared size-bounded dialog. Do not add a content minimum wider than the
+panel's padded interior. Genuinely wide content, such as a table, owns its own
+local scroller instead of making the entire dialog scroll horizontally.
+
+In `FormGrid` and `FormLayout`, direct `FormField` and `ToggleSwitch` items
+align their controls at the row's lower edge. A single control inside a
+`GridItem` follows the same rule. Labels may wrap without shifting adjacent
+switches up into the label row. Do not add per-module top margins or empty
+labels; single-column layouts must not retain a phantom label spacer.
+
+Credential editors resolve public reference labels when opened. A failed
+save displays its error inside the dialog and keeps the entered draft for an
+explicit retry. While a write is pending, repeated submission, edits, and
+dialog dismissal are disabled; no configured secret is read back from storage.
+
+The shared rich-text editor emits content changes only for actual document
+edits. Mounting, read-only changes, loading a saved value, and switching between
+visual and source inspection must preserve the controlled HTML without marking
+the owning page dirty. This is especially important for legacy Campaign HTML:
+merely visiting Template must not normalize it or require a save on leaving.
+The WYSIWYG lifecycle browser conformance covers both visual and legacy-source
+initial content, as well as genuine typing.
+
 Files and Mail are the first two external consumers of the layered
 server/credential/policy pattern. Their own repositories retain provider
 discovery, transport behavior, authorization, and migration evidence. Remaining
@@ -26,3 +76,15 @@ they are not reasons to add sibling-private behavior to Core.
 Raw JSON remains permitted only for diagnostics, expert inspection,
 interchange, or conflict evidence. It is not a primary Core configuration
 editor.
+
+New headed pages use `PageLayout`; full-canvas modules use `WorkspaceFrame`
+and, where applicable, `WorkspaceLayout`, so Core owns the frame and pane
+scrolling. Module CSS continues to own unequal domain content layout, never the
+shared page, workspace, toolbar, state, list, filter, metric, section, or graph
+chrome. Retired copies and module-local component definitions are rejected by
+`check-shared-webui-primitives.py`. That check also requires standard dialog
+widths to use `Dialog size` and keeps every remaining domain-specific width in
+a reviewed, decrease-only exception baseline. The companion layout check now
+has zero raw page-frame and zero raw workspace exceptions, discovers semantic
+consumers without a hand-maintained route list, requires semantic action bars
+on `WorkspaceFrame` routes, and rejects ad-hoc panel-header toolbars.

docs/LOCALIZATION_AND_HELP_QUALITY.md

@@ -0,0 +1,151 @@
+# Localization And Contextual Help Quality
+
+## Reference Language
+
+German (`de`) is GovOPlaN's first-class reference target. Every translation key
+used by a shipped WebUI must exist in German and English. German completeness is
+a release gate; English remains the source-code fallback language so existing
+literal labels and external developer APIs do not change semantics.
+
+New installations and tenants default to German. Existing system, tenant, and
+user preferences are preserved. The available-language and policy model can
+still select another default or disable a package at the relevant scope.
+
+Explicit high-risk help content and browser acceptance are tracked in
+[Core #284](https://git.add-ideas.de/GovOPlaN/govoplan-core/issues/284).
+
+The platform inventory recognizes both inline locale objects and generated
+catalogs declared as `const de` / `const en`. Its strict mode requires both
+locales and reports `de` explicitly as the reference locale.
+
+## Structured Documentation Localization
+
+`DocumentationTopic.translations` continues to own localized title, summary,
+and body prose. Topics whose metadata contains rendered prose opt into the
+separate `structured_translation_version="1"` contract and provide a complete
+same-shape value for each translated metadata key in
+`structured_translations`. Version 1 covers workflow prerequisites, steps,
+outcome, result and verification; reference fields; limitations, constraints,
+consequences and consequence classes; and the other rendered explanation
+fields declared by Core.
+
+The registry rejects an unversioned translation, an unsupported contract
+version, missing structured keys, changed object keys or list lengths, empty
+translated strings, and changed non-text values. Stable field IDs, routes,
+permission scopes, and other technical leaves therefore remain structurally
+bound to the source metadata. The Docs module overlays only a validated locale
+at response time and reports the selected structured locale separately from the
+title/body locale. Missing structured translations fall back to source content
+and remain visible in public coverage until the owning module adopts the
+contract.
+
+## Help Resolution
+
+Every focusable field and action receives a stable derived F1 identity from the
+shared shell, even when the component has no dedicated help text. Resolution
+falls back from field/action to dialog or page and then to the module's visible
+documentation baseline.
+
+Backend manifests publish explicit topic associations first. Core additionally
+associates declared route, navigation, settings, and View surface IDs with the
+module's static user or administrator documentation baseline. Feature modules
+should still add exact `metadata.help_contexts` entries for consequential,
+unfamiliar, policy-controlled, destructive, security-sensitive, or legally
+meaningful fields and actions.
+
+The shared retention-policy editor exposes explicit contexts for each stored
+data category, audit-detail control, lower-level override switch, target
+selector, reload, and save action. The Policy module owns the matching German
+administrator guidance. Retention execution surfaces use separate contexts for
+dry-run, destructive apply, confirmation, and outcome review so F1 opens the
+consequence and recovery guidance closest to the focused control.
+Shared controls may set `helpModuleId` when their documentation owner differs
+from the containing page; the retention editor uses this to resolve Policy help
+from both administration and Campaign surfaces.
+
+The shared reusable-credential manager keeps Access as its documentation owner
+and publishes exact contexts for credential kind, secret replacement/removal,
+module and server restrictions, lower-scope visibility, activation, save, and
+irreversible deletion. This ensures F1 explains secret custody and the effect on
+dependent connections from system, tenant, group, user, and personal surfaces.
+
+The source inventory treats literal `helpContextId` and
+`data-help-context-id` declarations as authored help associations, including a
+native control nested in `FormField`. Dynamic context expressions remain
+separate evidence and generic derived fallbacks remain in the richer-help
+candidate queue.
+
+The same inventory classifies controls whose labels, identities, component
+context, or explicit `data-help-risk` indicate authority, credentials,
+disclosure, encryption, external effects, irreversible changes, policy, or
+retention. These controls require an exact context rather than relying only on
+page fallback. Reviewed false positives carry
+`data-help-risk-reviewed="standard"`. Invalid risk classes and any increase
+above the versioned `tools/inventory/high-risk-help-baseline.json` ceiling fail
+strict declaration checks; the ceiling is lowered as the finite queue is
+resolved. Password fields and their generator dialog propagate the owning
+field's context so shared credential controls never invent a Core-owned topic.
+
+The generated `help_review_candidates` list is therefore a content-depth queue,
+not a list of controls on which F1 cannot work. It should prioritize:
+
+1. effect, deletion, delivery, retention, disclosure, encryption, and recovery;
+2. identity, representation, mandate, institutional context, and purpose;
+3. valid-time versus recorded-time selection;
+4. provider authority, synchronization, conflict, and outcome unknown;
+5. fields whose consequences are not evident from their label.
+
+The shared browser conformance journey mounts the production Help menu and
+resolver. It proves that F1 uses the focused control rather than only the page,
+maps an exact retention action to Policy-owned administrator documentation,
+retains the page context as fallback for derived actions, exposes an accessible
+modal at narrow widths, closes with Escape, and restores focus to the triggering
+control. Module journeys should add their own exact high-risk mappings; they do
+not need to reimplement the keyboard or dialog mechanics.
+
+The same conformance suite mounts the production Forms Runtime self-service and
+assisted Anwohnerparkausweis surfaces with German module translations. Desktop
+and mobile runs traverse native controls by keyboard, inspect accessible names
+and landmarks, run WCAG 2.1 A/AA automation, verify responsive overflow, and
+retain independent per-field assisted provenance. Physical assistive-technology
+spot checks remain release evidence rather than being represented as browser
+automation.
+
+## Verification
+
+```bash
+cd /mnt/DATA/git/govoplan
+/mnt/DATA/git/govoplan/.venv/bin/python \
+  tools/inventory/platform-interface-inventory.py \
+  --strict --strict-declarations --strict-endpoints
+```
+
+The check must report:
+
+- reference locale `de` present and complete;
+- no used key missing from `de` or `en`;
+- every field has a resolvable F1 context;
+- no duplicate stable IDs;
+- no undeclared public WebUI surface;
+- no stale runtime route or endpoint declaration.
+- no invalid high-risk help annotation or regression above the recorded
+  exact-context debt ceiling.
+
+Browser acceptance is part of the focused workspace gate and can be run alone:
+
+```bash
+cd /mnt/DATA/git/govoplan-core/webui
+npm run test:conformance
+```
+
+For shared component regressions, `npm run test:components -- page-layout
+documentation-help layout-primitives` compiles the component test configuration
+once and executes the selected suites. With no suite names it runs every shared
+component suite. Existing single-suite aliases such as `npm run test:page-layout`
+remain supported. Every invocation owns an ignored, uniquely named temporary
+output directory and removes it on completion, failure, or handled interruption;
+parallel invocations cannot delete each other's compiled tests. A forcibly killed
+process may leave its own ignored directory behind. The launcher itself is
+checked with `npm run test:component-runner` without compiling the application or
+starting a server. Browser conformance remains a separate verification layer;
+passing component tests does not constitute a complete module review.

docs/MODULE_ARCHITECTURE.md

@@ -16,7 +16,7 @@ The experimental remote WebUI bundle loading design is tracked in
 The cross-product semantic layers, source-authority modes, and candidate
 Mandates, Services, Parties, and Decisions boundaries are canonical in the
 meta repository's
-[`INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md`](../../govoplan/docs/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md).
+[`INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md`](../../govoplan/docs/architecture/INSTITUTIONAL_GOVERNANCE_TARGET_ARCHITECTURE.md).
 
 ## Layer Model
 
@@ -128,6 +128,8 @@ The following contracts are the baseline API that modules can rely on:
 - bounded reference-option search provider contract
 - single-tenant and optional batched tenant summary provider contracts
 - tenant delete-veto provider contract
+- provider-neutral tenant-erasure preview, step, idempotency, and
+  reconciliation contracts in `govoplan_core.core.tenant_erasure`
 - WebUI module contribution contract
 - navigation metadata contract
 - command/event envelope contract
@@ -149,6 +151,17 @@ Destructive tenant lifecycle planning deliberately continues to use the
 single-tenant path so it invokes every registered provider for the target
 tenant, independent of ordinary list-page projections.
 
+Governed populated-tenant erasure is separate from ordinary delete vetoes.
+Modules contribute `tenancy.erasure_provider.<module_id>` capabilities with a
+bounded resource inventory, explicit erase/retain/legal-hold/external/key/
+backup dispositions, ordered destructive warnings, idempotent step execution,
+and reconciliation. The collector fails closed when a provider is invalid or
+fails. A module with nonzero tenant summary counts and no erasure capability is
+reported as unsupported and blocks execution; modules with neither contract
+are explicitly projected as outside tenant-persistence scope. Provider
+evidence contains counts and stable references only and must never contain
+secrets or erased subject data.
+
 This list is the Milestone A kernel-contract freeze baseline. New module work
 may extend the kernel by adding explicit contracts, but existing contracts must
 remain source-compatible through the 0.1.x split line unless a migration shim
@@ -213,8 +226,29 @@ Other stable runtime capabilities currently include:
   `calendar.externalProfiles`
 - `poll.scheduling`
 - `notifications.dispatch`
+- `application_status.projection`
+- `payments.requests`
 - `workflow.definitionContributions` and `workflow.runtimeWorker`
 
+`calendar.scheduling` keeps workflow modules independent of Calendar-owned
+models and transport adapters. Consumers may create tentative events, promote
+the selected event in place, and release unused events idempotently. The
+provider returns bounded external-delivery and outbox references so consumers
+can retain retry state without copying Calendar's synchronization internals.
+
+`application_status.projection` lets a presentation module resolve the tenant
+and display or request access to an owner-supplied, deliberately bounded
+applicant-status view. The provider retains policy, authorization, token, and
+record ownership; consumers must not query provider tables or enlarge the
+projection.
+
+`payments.requests` carries replay-safe payment obligations and evidence-bound
+manual reconciliation across module boundaries. Procedure modules identify the
+source Case or Workflow in the command and retain the returned payment ID;
+Payments remains authoritative for amount, currency, state, transaction
+reference, and reconciliation evidence. Ledger, invoice, and external payment
+providers remain separate follow-on contracts.
+
 The provider-neutral `idm.relationships` contract carries tenant-scoped typed
 groups, effective-dated identity relationships, and explicit membership
 decisions. It deliberately does not expose IDM persistence models or imply an
@@ -305,6 +339,8 @@ contract checks, are:
 - `files.access`, `files.campaign_attachments`
 - `mail.campaign_delivery`
 - `notifications.dispatch`
+- `application_status.projection`
+- `payments.requests`
 - `poll.availability_matrix`, `poll.option_selection`,
   `poll.response_collection`, `poll.signed_participation`,
   `poll.workflow_context`
@@ -1007,6 +1043,49 @@ Any future exception is extraction debt and must be temporary, documented in the
 script with a reason, and removed when a capability/API/event contract replaces
 it.
 
+## Product Surface Contributions
+
+`FrontendModule.product_surfaces` is the versioned product-composition contract
+for stable identities that may have one or more technical owners. A contribution
+declares contract version 1, a product identity, common label/icon/description,
+stable entry path, owner route and View surfaces, supported task/reader/admin/
+operator presentations, authorization requirements, capabilities, search
+sources, help contexts, documentation topics, migration aliases, and standard
+unavailable/degraded explanations.
+
+Core validates every reference against the owning manifest. Contributors that
+share an identity must agree on its common product metadata and entry path;
+entry and alias paths cannot belong to another product identity. The WebUI
+composes valid owners by product id, filters them through authorization and the
+effective View, and resolves the stable entry or migration alias to the first
+available owner route. It emits `govoplan:product-surface-route-resolved` before
+the redirect so migration telemetry can observe alias use without making the
+technical module part of the ordinary label.
+
+The shell projects every authorized, View-visible owner route with a product
+contribution into one stable product navigation item. The product label and
+entry path replace package topology in the primary rail; every contributing
+owner path still marks that item active. `All available tools` is a collapsed,
+permission-derived catalogue built independently of the active View, so a
+focused workflow cannot remove the explicit escape. It may reveal an
+authorized owner route that a View omitted, but never an unauthorized route.
+Navigation visibility preferences do not delete catalogue entries, and the
+original owner routes remain compatible deep links.
+
+The initial promoted destinations are `work.items` at `/work`,
+`meetings.calendar` at `/agenda`, `communication.messages` at `/messages`
+(with `/inbox` as an alias), and `records.files` at `/documents`. Their labels
+and availability language are centralized in Core while Tasks, Calendar,
+Mail/Postbox, and Files retain route, command, search, help, documentation,
+authorization, and data ownership.
+
+Use `ProductAvailabilityState` for unavailable and degraded outcomes. The
+ordinary state explains the attempted outcome, consequence, recovery path and
+responsible role. Exact module, capability, provider and correlation values may
+be supplied as a collapsed technical detail; they are not the primary error.
+The state is presentation only and never grants authority or changes provider
+health.
+
 ## Boundary Decision Register
 
 These durable decisions close older exploratory core issues. Implementation
@@ -1377,8 +1456,10 @@ The package install-plan API records operator intent only:
 - `GET /api/v1/admin/system/modules/package-catalog` reads approved package
   references from `GOVOPLAN_MODULE_PACKAGE_CATALOG` so operators can add known
   module refs to the install plan without typing them manually. The endpoint
-  also reports catalog validity, channel, signature, trust state, and the
-  configured path.
+  also reports catalog validity, channel, signature, trust state, source and
+  artifact provenance, release availability, configuration requirements, and
+  per-entry compatibility/blocker state. Withdrawn entries are visible for
+  diagnosis but cannot be planned.
 - `POST /api/v1/admin/system/modules/install-plan/catalog/{module_id}` saves
   a planned install or update row from a validated catalog entry. Installed
   modules are planned as updates. Catalog signature and approved-channel policy
@@ -1476,6 +1557,10 @@ The installer preflight is intentionally conservative:
 - the `shared` state profile blocks in-place package mutation; clustered
   installations must roll one verified immutable module composition across all
   replicas;
+- official runtime images carry the full verified package profile, while the
+  desired module graph controls activation and tenant/View/Policy contracts
+  control availability and presentation; package lifecycle must not be reused
+  as a tenant or user visibility switch;
 - installed module manifests must be compatible with the supported manifest
   contract and current core version;
 - uninstalling `tenancy`, `access`, or `admin` is blocked;
@@ -1550,6 +1635,16 @@ Unsigned/unhashed remote bundles are skipped. This keeps remote loading a
 controlled deployment option rather than a replacement for release package
 builds.
 
+A failed local WebUI package import receives one automatic retry after 250 ms.
+Descriptor validation still fails closed; it is not bypassed by the retry.
+If an enabled local module still cannot load, the signed-in shell warns that its
+screens and integrations may be unavailable and identifies the module. This is
+a loading failure, not an uninstall. Save other drafts before manually reloading
+the page; there is no automatic page reload or persistent retry loop. A verified
+remote fallback that successfully loads clears that module's warning. Packages
+absent from the optional build graph remain absent, and effective View filtering
+continues to control which loaded UI capabilities are exposed.
+
 ## Maintenance Mode
 
 Maintenance mode is the required operating state for package install/uninstall
@@ -1592,6 +1687,17 @@ URLs never contain credentials; only credential-envelope references cross the
 contract. Provider-specific details belong in sanitized provenance rather than
 in a shared domain schema.
 
+## Semantic Documentation Subject Contract
+
+Optional modules expose configured artifacts that can be documented through
+the module-scoped `documentation.semantic_subjects.<module_id>` capability.
+Core supplies stable tenant-scoped references, typed nested anchors, safe
+localized descriptors, revision/fingerprint review signals, and explicit
+availability states. Providers remain responsible for authorization and do not
+expose configuration payloads or credentials. Docs discovers the capability
+and owns authored content; it does not import feature internals. See
+`SEMANTIC_DOCUMENTATION_SUBJECTS.md` for the contract and adoption rules.
+
 ## Build And Verification
 
 Backend verification from core:

docs/NAVIGATION_LAYOUT_CONTRACT.md

@@ -0,0 +1,96 @@
+# Shared navigation layout contract
+
+Core owns `NavigationPreferenceEditor`, ordered layout resolution, and rail
+rendering. Admin, Tenancy, personal Settings and Views reuse this editor. They
+own loading, authorization, Save, Reload and dirty-state guards; the editor
+emits a draft only after a real edit. A drag onto the same position, keyboard
+pickup/drop without movement, and opening inherited settings do not save or
+create an override.
+
+## Standard layout / Standardanordnung
+
+No configuration is required to enable sections. Installed modules contribute
+the standard product-area membership, labels, and order: Work; Services and
+cases; Records and documents; Communication; Meetings and decisions; Data and
+assurance; People and responsibility. The dashboard stays first; unclassified
+destinations remain available under More tools. Only nonempty, authorized and
+View-visible groups appear. Composed product entries keep their placement
+owner's directly declared area. If none exists, the navigation aliases of
+other authorized owners preserve area membership. These are live defaults,
+not a saved copy that needs replacing whenever optional modules change.
+
+The editor explains whether the current level is inherited, custom grouped, or
+explicitly flat. To restore defaults, choose **Use inherited layout** and Save
+on the owning page: personal Settings inherits the tenant, tenant settings
+inherit the system, and system settings inherit the standard product areas.
+An explicitly flat personal, tenant, system, or View configuration is never
+silently replaced with standard groups. A View may also explicitly select flat
+navigation. Reset the responsible override rather than changing permissions.
+
+Abschnitte sind ohne zusätzliche Konfiguration aktiv. Die Standardanordnung
+gliedert verfügbare Module in Arbeit; Leistungen und Vorgänge; Akten und
+Dokumente; Kommunikation; Termine und Entscheidungen; Daten und
+Qualitätssicherung; Personen und Verantwortung. Das Dashboard steht davor,
+nicht zugeordnete Ziele bleiben unter Weitere Werkzeuge erreichbar. Leere oder
+nicht zugängliche Gruppen erscheinen nicht. Der gemeinsame Editor zeigt an,
+ob eine geerbte, eigene gruppierte oder ausdrücklich ungegliederte Anordnung
+vorliegt. **Geerbte Anordnung verwenden** und anschließendes Speichern entfernt
+nur die Anpassung dieser Ebene: persönlich → Mandant → System →
+Standardanordnung. Bewusst gespeicherte ungegliederte Anordnungen bleiben
+erhalten. Diese Darstellung erteilt keine Berechtigungen und ändert keine Daten.
+
+## Stored document and precedence
+
+The version-1 navigation document retains `order`, `hidden` and `locked` and
+adds optional `separators`, each containing a stable `separator:`-prefixed ID
+and an optional plain-text label of at most 120 characters. Separator IDs and
+module navigation IDs occupy the same `order` list. Separators are presentation
+metadata and never become routes, modules, permissions or authorized surfaces.
+
+Omitting `separators` or using null preserves inherited grouping. An explicit
+empty array removes grouping. Resetting the entire navigation document to null
+removes that scope's override. Existing order-only documents remain readable;
+the editor materializes group markers into a draft only when edited. Unknown
+optional-module order IDs remain stored when currently visible items move, so
+uninstalling or temporarily disabling a module does not destroy its preference.
+
+User order and visibility take precedence over tenant and system preferences.
+System and tenant visibility locks accumulate; lower scopes cannot hide those
+destinations, but may move them. Views may supply a navigation presentation
+inside the already authorized and View-filtered destination set. An explicit
+personal order/layout or visibility preference takes precedence over that
+presentation, not over authorization or the View's surface restrictions.
+Views cannot introduce locks. When multiple modules contribute one product
+entry, it inherits the earliest effective rail position/section and all
+authorized contributors' locks; this does not change operational route
+selection. An owner alias in View layout refers to that composed entry.
+
+## Interaction and reuse
+
+Drag the handle to move either a module or separator before/after another row.
+The handle also supports Space to pick up, arrow keys to move, Enter to drop,
+and Escape to restore the pre-drag draft. Up/down buttons offer the same moves.
+Add module restores an available hidden entry; Remove only hides navigation,
+never uninstalls a module or deletes records. Add separator inserts a new
+optional group label. Remove separator changes grouping only.
+
+Expanded rails display group labels without divider lines. Collapsed rails
+replace these labels with horizontal group dividers; empty groups are not
+rendered. Both modes use the same resolved order.
+The editor receives product-area metadata to show inherited grouping and uses
+container-responsive rows rather than a fixed dialog/page width. Its English
+and German labels load with the editor, not the initial shell bundle.
+Give the ordered list a full-span `GridItem` when a settings page contains
+multiple cards; do not squeeze the entire editor into an otherwise half-empty
+two-column settings grid. Central spacing tokens provide real row padding and
+separation at both wide and narrow sizes, covered by computed-style assertions.
+
+## Verification
+
+Core navigation unit tests and HTTP settings/profile tests cover persistence,
+separator inheritance, explicit flat layouts, reset and locks. Module-capability
+tests cover View aliases, composed destinations and personal precedence. The
+browser conformance suite tests all four editor scopes, pointer/keyboard moves,
+no-op cleanliness, unavailable-module preservation, collapsed dividers and a
+German narrow read-only layout. Use the same shared component for future
+navigation-definition surfaces rather than implementing another sortable list.

docs/PAGE_LAYOUT_USAGE_GUIDELINES.md

@@ -0,0 +1,251 @@
+# Page Layout and Action Guidelines
+
+This document defines the binding composition grammar for headed GovOPlaN
+pages. Core owns the reusable anatomy; each module owns its domain actions,
+wording, authorization, consequences, and data state.
+
+The cross-product [UI design principles](UI_DESIGN_PRINCIPLES.md) define the
+display-first editing model, heading-adjacent help, and module-review process.
+Normal overview/detail surfaces show readable facts; edit coherent settings in
+scoped dialogs. Use an explicit editor mode for a justified broad-editing task,
+not merely because a page also offers filters or contains an edit dialog (UI-02).
+
+## Required Page Frame
+
+- Use `PageLayout` for every headed standalone, workspace, or embedded page.
+- Declare exactly one semantic `archetype`; do not infer page intent from the
+  `mode`, which controls geometry and scroll ownership only.
+- Use `WorkspaceFrame` for a full-height module surface and
+  `WorkspaceLayout` only where navigation/content or list/detail panes are
+  genuinely part of the interaction.
+- Put page-wide feedback in `PageLayout` notices. Use `DismissibleAlert` for a
+  recoverable warning or failure and `StatePanel` when the entire surface is
+  loading, empty, unavailable, or blocked.
+- Do not reproduce shared page padding, heading, toolbar, form-grid, section,
+  table, dialog, or breakpoint CSS in a module.
+
+## Product Side Rail
+
+Module manifests contribute stable navigation surface identifiers, labels,
+paths, icons, and default order. Core owns the side-rail composition and the
+shared `NavigationPreferenceEditor`; modules must not fork this editor or
+persist their own rail ordering.
+
+Navigation preferences are layered in this order: module defaults, system,
+tenant, then user. Each higher layer may reorder or change visibility. System
+and tenant administrators may lock an entry visible; a lower layer can still
+move that entry, but cannot hide it. Personal preferences cannot create locks.
+An unset preference inherits the complete lower layer, while “Use inherited
+order” removes the current layer rather than copying its values. Unknown item
+identifiers remain harmless so uninstalling, disabling, or later reinstalling
+a module does not corrupt the rail.
+
+The platform module response projects module, system, and tenant layer states
+alongside the effective user state. Editors must initialize from the layer
+immediately below the scope they edit, so a system or tenant administrator's
+personal preference is never promoted accidentally. Preference saves refresh
+the platform module projection. View policy, permissions, and tenant module
+entitlements remain independent final visibility gates; changing rail
+preferences never grants access.
+
+## Semantic Page Archetypes
+
+| Archetype | Use when |
+| --- | --- |
+| `overview` | The page summarizes health, metrics, or several peer areas without owning one primary collection or draft. |
+| `collection` | The primary object is a searchable/listable collection and Create, when available, applies to that collection. |
+| `detail` | The page primarily presents one record, report, or immutable projection. |
+| `editor` | The page owns one explicit draft with Save and Discard behavior. |
+| `workspace` | The page coordinates several panes, stages, or task-local operations that cannot honestly be reduced to one record or draft. |
+
+The archetype remains stable for the current interaction. A page may switch
+from `overview` to `editor` when the user explicitly enters configuration
+mode. It must not call a page an editor merely because a dialog or an inline
+filter is editable.
+
+## Page Action Rules
+
+Pass one `PageActionBar` to the `PageLayout` `actions` slot. Full-canvas
+workspaces use the same contract through `WorkspaceActionBar`, with an explicit
+`workspace`, `collection-pane`, `detail-pane`, or `editor-pane` scope. The
+variant makes the surface's intent inspectable and preserves the same keyboard
+and visual order across modules. `ActionToolbar` remains the lower-level
+component for section-local controls; it is not a substitute for a semantic
+page or pane action bar.
+
+| Page kind | Leading group | Trailing group |
+| --- | --- | --- |
+| Overview | Context | Reload when refreshable, then ordinary primary actions |
+| Collection | Collection context such as export | Reload when refreshable, then Create at the far right |
+| Detail | Object context | Reload when refreshable, ordinary primary actions, then a separated destructive group |
+| Editor | Context | Dirty state, Reload if distinctly safe, ordinary primary actions, separated destructive actions, Discard, then Save at the far right |
+| Workspace | Task context | Reload when refreshable, ordinary primary actions, then a separated destructive group |
+
+Documentation is associated with text, not an action-group slot (UI-01). Put
+`DocumentationHelpLink` in the heading component's `titleHelp`, use `TextWithHelp`
+for existing contextual words, and retain `FormField`/`FieldLabel` documentation
+beside field labels. Full-canvas `WorkspaceActionBar` surfaces can supply `title`
+and `titleHelp` together on the leading side. Do not duplicate a heading or use
+`helpAction` for a detached documentation icon. See the shared principle for
+card/dialog accessibility and examples.
+
+Reload and Create belong to the same right-aligned group, in that order. A
+collection-wide toolbar stays above its workspace, not inside the left tree or
+conditionally inside an editor. Changing selection or opening an editor must
+not remove it. Permission-blocked creation remains visible with an explanation.
+On narrow screens the trailing group wraps while retaining right alignment and
+the same DOM/keyboard order.
+
+An editor's **Cancel/Close** action is not the same as **Reset/Discard changes**.
+When it exits an editing mode, set `discardAction.behavior` to `"exit"`: it stays
+enabled for a clean draft, while Save remains disabled until there are changes.
+The default `"reset"` behavior still requires changes. Both stay blocked during
+an in-flight save; the owning page must use the shared discard confirmation before
+exiting a dirty draft. Never require a meaningless edit just to leave configuration.
+
+Deutsch: Abbrechen oder Schließen beendet den Bearbeitungsmodus auch ohne
+Änderungen (`discardAction.behavior="exit"`). Zurücksetzen setzt dagegen einen
+geänderten Entwurf voraus. Beim Abbrechen eines geänderten Entwurfs vor dem
+Verwerfen nachfragen; eine laufende Speicherung bleibt geschützt.
+
+Use `Card bodyLayout="table"` for table surfaces, including tables wrapped by
+`LoadingFrame`. This removes body padding explicitly, without relying on the
+number of children or negative margins. Place any meaningful explanation or
+warning in a padded `ContentSection`; do not add a redundant tagline to every
+table. Use `ContentGrid` for sibling cards so spacing does not depend on fragments.
+
+The shared stylesheet also recognizes existing table-only card bodies, either
+directly or through `LoadingFrame`; loading must not add an inset, expand the
+overlay outside the card, or change the table's measured width. Keep the table
+at 100% of its container. Do not reintroduce module-local negative margins,
+expanded `calc()` widths, or padding overrides to make a table fit. Mixed
+content cards retain their normal padding unless they explicitly declare the
+table layout. The browser conformance matrix checks all four card edges,
+loading/collapse transitions, scrolling, and row actions at desktop and mobile
+widths.
+
+Deutsch: Tabellen belegen den Karteninhalt ohne zusätzlichen Innenabstand,
+auch während des Ladens. Neue Tabellenkarten verwenden ausdrücklich
+`bodyLayout="table"`; ergänzende Hinweise erhalten ihren eigenen Innenabstand.
+Negative Ränder oder modulbezogene Breitenkorrekturen sind nicht nötig.
+Karten mit gemischtem Text- oder Formularinhalt behalten ihre normalen Abstände.
+
+Use `MultiSelectFilter` for standalone list facets. It and DataGrid share the
+same checkbox body and Select all / Deselect all behavior. `null` means no
+restriction, `[]` means no matches, and multiple values mean OR within a facet.
+Apply remote filters before server pagination/limits and discard stale reads.
+Do not replace this with rows of toggles or implement a second checkbox menu.
+The dropdown's body portal escapes clipped containers. Inside Core dialogs it
+joins the existing dialog stack: Tab/Shift+Tab stay in the filter, Space toggles
+the focused checkbox, and Escape closes only the filter and restores its
+trigger. Long option labels wrap without widening the popup.
+
+Keep facet definitions, URL serialization and request cancellation in one owning
+module adapter when the same search appears on a page and in an overlay. Do not
+translate an explicit empty selection into an unrestricted backend query. Keep
+legacy API meanings at the adapter boundary; retain scope and unrelated URL
+parameters when clearing filters. A query, context, account or tenant change
+invalidates both initial and cursor requests, including results still visible
+during a debounce interval.
+
+Explorer workspaces keep collection commands in a persistent header. Files
+uses Reload, Create folder, then primary Upload; frequent selected-item actions
+stay near the list. Group less-common selection and connection operations in
+labelled domain dialogs using `Dialog`, `FormSection` and shared action bars,
+with an explicit destructive section. Do not move an overloaded toolbar into
+another ungrouped row. Mail's read-only workspace has one Reload for its current
+profile, folder, index and preview; narrower refreshes belong in Mailbox tools.
+Do not invent a New or Save button on a workspace that owns neither workflow.
+Reload must not become import, synchronization, delivery or another mutation.
+Explicit Reload reads must bypass short-lived client response reuse (for
+example, pass `cache: "no-store"` through the owning read API), including each
+page of a refreshed listing. Routine navigation may retain normal deduplication.
+Conformance must observe a fresh request, not just an enabled Reload button.
+
+Tree icons/disclosure controls expand and collapse; labels select. A module's
+`ExplorerTree.onOpen` must not toggle expansion. Use occurrence-specific node
+IDs when the same semantic record appears in multiple branches; selection and
+ancestor expansion must follow the clicked occurrence, not every copy.
+
+Reload means re-fetch or re-evaluate the current surface. A page declaring
+`refreshable` must provide it, and a non-refreshable page must not use Reload as
+a synonym for Cancel, Reset, or Discard. Reload never silently destroys a dirty
+draft. Create is a collection-wide action and is not duplicated in a
+persistent side panel. Save is present only where the page owns an editable
+draft; a read-only detail page must not display a disabled or inert Save merely
+to fill the slot.
+
+Editor bars always keep Discard and Save visible. Their required `state`
+projection is one of `clean`, `dirty`, `invalid`, `saving`, `save-failed`, or
+`conflict`, and the central component announces it through a live status label.
+Clean states disable Save and reset-style Discard, but keep exit-style Cancel
+available; saving disables both persistence actions. Invalid disables Save
+while retaining Discard. Failed saves and conflicts keep the draft recoverable
+and allow an authorized retry after the module has shown the owning error or
+conflict evidence. A module may add a more specific validation, policy, or
+permission blocker. The editor must register its draft with
+`useUnsavedDraftGuard` (or a shared hook that uses the same registration
+contract), so browser unload, route navigation, section changes, Reload, and
+the explicit Discard path cannot silently lose work.
+
+Reload is rendered by Core from a descriptor rather than passed as arbitrary
+button markup. It can project `current`, `stale`, `reloading`, or
+`reload-failed`; `loading` is the shorthand for `reloading`. A failed refresh
+must preserve usable loaded data, expose its stale/failure state, and leave
+Reload available for recovery. Reload goes through the same unsaved-navigation
+guard as route changes.
+
+Destructive page actions use `destructiveActions`; never put a danger action in
+`contextActions` or the ordinary primary group. Core renders a persistent
+visual and semantic boundary before this group. In an editor it precedes the
+Discard/Save pair, keeping Save in the final keyboard and visual position.
+
+`PageActionBar` controls non-editor placement and owns the standard editor
+persistence buttons. Other actions continue to use central
+`Button`, `IconButton`, or `TableActionGroup` components. When an action is
+visible but unavailable because of permission, target, policy, state, or
+validation, keep it in its stable slot and supply `disabledReason`. Do not
+silently hide a normally applicable action.
+
+## Forms and Dialogs
+
+- Compose forms from `FormLayout`/`FormGrid`, `FormSection`, and `FormField`.
+- Use `FieldLabel` through `FormField` for every field that is not genuinely
+  self-explanatory; record justified omissions in the owning UI ledger.
+- Use `Dialog`, `DialogForm`, `DialogSection`, and `DialogActions` for modal
+  work. A dialog can be domain-specific while its anatomy remains central.
+- Use `useUnsavedDraftGuard` for explicit Discard and guarded navigation on an
+  editable page or dialog.
+- Explain irreversible or operationally consequential actions before the
+  commit button, including reversibility and durable evidence.
+
+## Collections and Details
+
+- Use `FilterBar` for collection query controls and `DataGrid` for tabular
+  collections. Keep a single ordered `TableActionGroup` action set per table.
+- Use `MetricGrid`/`MetricCard` for summary measures, `Card` or
+  `ContentSection` for logical sections, and `DescriptionList` for labelled
+  facts.
+- Add a `MetricCard.drilldown` only when the displayed measure has a useful,
+  authorized underlying collection or detail. Name the destination explicitly
+  (for example, “Review failed deliveries”) and preserve the current scope and
+  filters in its `href` or action. The card itself remains non-interactive so
+  the action is visible and keyboard-predictable. Derived, privacy-suppressed,
+  non-enumerable, or purely informational aggregates remain plain metrics;
+  when an ordinarily available drill-down is temporarily blocked, keep its
+  action and provide `disabledReason`.
+- Preserve loaded data after a refresh failure and mark it stale; offer Reload
+  as the recovery action. Distinguish initial loading, empty, unavailable,
+  permission-blocked, conflict, success, and retry states.
+
+## Review Evidence
+
+Every new or changed page or workspace pane must have structural evidence for
+its frame, semantic archetype/scope and slot order, refresh declaration, shared
+component usage, stable disabled actions, dirty guard, destructive boundary,
+and module-owned help identity. Type checks enforce conditional Reload and
+editor persistence props. The product check discovers all consumers, rejects
+undeclared archetypes and `ActionToolbar` panel-header copies, and requires
+semantic actions for every `WorkspaceFrame` route. Browser conformance confirms
+keyboard order, lifecycle changes, accessibility, destructive separation,
+narrow wrapping, and screenshot geometry.

docs/POLICY_CONTRACTS.md

@@ -14,6 +14,7 @@ consistent while each module still owns its domain rules.
 | Governance defaults | `govoplan-admin` plus `govoplan-access` materializer | admin settings, governance template routes, access materialization capability | System governance can block tenant-local groups, roles, and API keys. |
 | Delegation and ownership policy | access/campaign/mail/files modules | capability checks and owner-scoped APIs | Source provenance should use this contract when policies become externally explainable. |
 | Definition governance | `govoplan-policy` | capability `policy.definitionGovernance` | Resolves view, edit, run/start, reuse, derive, and automate for system, tenant, group, and user Dataflow/Workflow definitions. |
+| Function assignment governance | `govoplan-policy` | capability `policy.functionAssignmentGovernance` | Returns current review steps, delegation depth/validity ceilings, and explicit timed-escalation targets consumed by IDM. |
 
 ## Policy Decision
 
@@ -126,6 +127,43 @@ When the capability is absent, modules must not silently emulate cross-scope
 inheritance. Their conservative fallback is limited to local tenant
 definitions and disables reuse, derivation, and automation.
 
+## Function Assignment Delegation And Escalation
+
+`FunctionAssignmentGovernanceDecision` is the versioned cross-module contract
+for request/grant review. In addition to the required holder, authority, and
+recipient steps, it returns `delegation_allowed`,
+`maximum_delegation_depth`, `maximum_delegated_validity_days`, and typed
+`FunctionAssignmentEscalationRule` entries. Each escalation entry binds one
+review step to an exact target function and timeout.
+
+The decision is a current ceiling, not durable authorization. IDM must recheck
+the complete assignment-source chain and all recorded decisions before final
+application. An elapsed timeout creates explicit state and evidence; it must
+never be interpreted as approval or as permission to silently substitute an
+approver. Missing providers, malformed rules, invalid chains, or tightened
+limits fail closed with an explainable reason.
+
+## Bounded Impact-Subject Providers
+
+Policy impact previews discover optional subject providers through capability
+names beginning with `policy.impactSubjects.`. The suffix is the stable
+provider ID; for example, Views contributes `policy.impactSubjects.views`.
+Providers implement `PolicyImpactSubjectProvider` and receive a
+`PolicyImpactPopulationRequest` containing the active tenant, policy family,
+an explicit selector, actor scopes, detail-disclosure decision, and a limit of
+at most 500. They return `PolicyImpactSubjectBatch` with unique opaque subject
+references and an explicit `complete`, `sampled`, `truncated`, or `unavailable`
+state. An unavailable batch must explain the gap, and a total may never be
+smaller than the returned subject count.
+
+Core does not scan module data or evaluate domain policy. The owning module
+selects and permission-filters its candidates; Policy compares the current and
+proposed decisions and controls response disclosure. A caller must select one
+or more provider populations explicitly. This preserves optional-module
+boundaries and prevents a seemingly harmless preview from becoming an
+unbounded platform query. Providers must not include credentials, secrets, or
+unfiltered cross-tenant labels in subject attributes.
+
 ## Frontend Contract
 
 Policy UIs must:
@@ -138,6 +176,9 @@ Policy UIs must:
   lower-level limit to `false`
 - avoid sending locked fields or re-enable attempts in save payloads
 - show inherited values separately from local overrides
+- require a current impact preview before enabling a governed high-impact save,
+  preserve its proposal hash on commit, and explain incomplete population
+  coverage rather than presenting unavailable providers as zero impact
 
 The core WebUI helper `privacyRetentionParentAllowsField()` centralizes the
 field-lock decision used by the retention editor and its lightweight module

docs/POSTBOX_E2EE_ARCHITECTURE.md

@@ -1,9 +1,12 @@
 # Postbox End-To-End Encryption Architecture
 
-This document records the strategic encryption target for GovOPlaN postboxes.
-It does not require the first postbox implementation to ship full E2EE, but it
-defines the architecture so early data models and APIs do not make the stronger
-model impossible.
+This document records the encryption boundary for GovOPlaN postboxes. Postbox
+now implements the server-side contracts for three selectable profiles:
+unencrypted content, institution-managed server envelopes, and externally
+produced E2EE envelopes. The E2EE contract is operational—the server rejects
+plaintext and retains ciphertext, signed manifests, wrapped keys, and digest
+evidence—but a reviewed browser/device client and private-key custody provider
+remain separately deployed responsibilities.
 
 The core principle is that a postbox can become a trusted administrative
 communication channel without requiring the server to see plaintext content.
@@ -35,6 +38,54 @@ Algorithm choices should remain replaceable behind a crypto profile. The first
 profile should prefer standard, reviewed primitives such as HPKE for key
 wrapping and AEAD encryption for content.
 
+## Product Profiles And Default
+
+The content-protection policy is configurable per exact Postbox or immutable
+template revision:
+
+- `server_envelope_v1` is the recommended default. An institution-selected
+  Encryption vault controls server-readable envelopes and their migration
+  evidence. It is not end-to-end encryption.
+- `external_e2ee_v1` is server-blind. An approved client or producer supplies
+  the ciphertext reference, signed manifest, wrapped recipient keys, key epoch,
+  and SHA-256 plaintext digest. GovOPlaN has no private key that can decrypt it.
+- `plaintext_v1` stores clear content for institutions that explicitly choose
+  that boundary.
+
+Operational metadata—including subject, routing, participants,
+classifications, timestamps, attachment references, receipts, and retention
+state—remains visible under every profile. Administrators therefore choose a
+content-protection boundary, not a metadata-anonymity profile.
+
+The standard policy grants new incumbents history since assignment, uses key
+rewrapping for ordinary rotation and content re-encryption after compromise,
+requires two-person institutional recovery and dual-control hand-over,
+emergency, export, and destruction, requires strong external identity, and
+limits vacancy escalation to metadata. Deployments may select other policy
+values rather than inheriting a decision from GovOPlaN.
+
+## Governed Profile Changes
+
+A profile transition applies to new messages immediately and increments the
+Postbox key epoch. Retained history can remain under the previous profile or be
+migrated. The transition ledger records source and target profiles/vaults,
+authority route, consent and key-holder evidence, quorum, reason, immutable
+configuration snapshot, per-message source and target digest, and outcome.
+
+Plaintext and managed-envelope migrations can use the server-side Encryption
+capability. Managed decrypt, export, and re-encryption operations also create
+Encryption migration records so old envelopes are disposed of through the
+governed provider contract. Any transition to or from E2EE pauses each retained
+message for an approved client transform. The client must return plaintext or
+ciphertext as appropriate, plus evidence and the original content digest;
+Postbox verifies digest continuity before changing the stored representation.
+Leaving E2EE requires user-consent evidence, while changing managed history
+requires institutional key-holder evidence. Dual control can require both.
+
+This transition mechanism cannot revoke plaintext already decrypted, copied,
+printed, or exported. Administrators must explicitly acknowledge that residual
+disclosure before a transition is accepted.
+
 ## Identity And Device Keys
 
 The platform should distinguish:

docs/PUBLIC_SECTOR_INTEGRATION_STRATEGY.md

@@ -141,6 +141,73 @@ connector or module issue.
 - Owner/priority: `govoplan-mail`, `govoplan-calendar`,
   `govoplan-connectors`, Wave 1/2.
 
+#### Collaboration-suite boundary and hand-offs
+
+Collaboration remains connector-first. The product named below never changes
+which GovOPlaN module owns the administrative meaning of the work:
+
+| External family | Initial posture | GovOPlaN semantic owner | Connector-owned boundary |
+| --- | --- | --- | --- |
+| Collabora Online, OnlyOffice, Nextcloud Office | Link an externally edited document and its editing session; import a governed rendition only when required | DMS owns document/version, lock, review, approval, retention, and collaboration-session evidence; Files owns stored bytes | Discovery, endpoint health, WOPI/vendor session exchange, callbacks, and provider object references |
+| Matrix, Mattermost, Rocket.Chat, Nextcloud Talk | Create or link a room/thread for a governed work context; do not mirror all conversation history by default | The initiating Case, Workflow, or Task owns the work-context link and disposition; DMS/Records own retained evidence deliberately captured from it | Room/thread creation, membership synchronization, webhook/event normalization, and stable external links |
+| Jitsi and BigBlueButton | Provision or link a conference for an existing appointment/event | Appointments owns booking intent; Calendar owns event, attendee, invitation, and time semantics | Conference provisioning, join/moderator references, provider lifecycle, and bounded attendance/result callbacks |
+| OpenProject and comparable project suites | Link first, then publish or synchronize selected work packages | Tasks owns GovOPlaN task state; Workflow owns orchestration; Cases own case state and evidence references | Project/work-package lookup, publish/synchronize transport, webhooks, version tokens, and external URLs |
+| Cross-suite activity streams | Consume normalized, bounded events only for an authorized work context | The receiving module decides whether an event changes state or becomes evidence; Audit records the GovOPlaN operation | Provider subscriptions, cursor/checkpoint handling, signature validation, event normalization, and replay protection |
+
+Native collaboration behavior is justified only when GovOPlaN must own the
+semantic state, authorization decision, audit evidence, retention/legal-hold
+rule, or configuration-package fragment. Endpoint profiles, tokens, health,
+protocol clients, provider IDs, retries, and webhook transport remain in
+Connectors (or the owning protocol connector). A feature module consumes a
+Core capability/DTO and must still start and fail explicitly when that optional
+connector is absent; it never imports a provider client.
+
+The minimum hand-off sequences are:
+
+1. **Appointment to conference:** Appointments confirms the booking intent;
+   Calendar creates or updates the event and invitations; an optional
+   conference connector provisions the room idempotently and returns an
+   opaque join reference. Calendar stores that reference with the event, not
+   the provider credential.
+2. **Case or Workflow to collaborative document:** the initiating module asks
+   DMS for a governed document/session; DMS requests an optional office-suite
+   connector session and retains version, lock, approval, and callback
+   evidence. The Case/Workflow keeps only the DMS reference.
+3. **Case, Workflow, or Task to chat:** the semantic owner requests a room or
+   thread with an idempotency key and bounded membership intent. The connector
+   returns an external reference; capturing messages as evidence requires an
+   explicit DMS/Records action and policy decision.
+4. **Task or Workflow to project suite:** Tasks supplies the task payload and
+   Workflow supplies correlation; the OpenProject connector publishes or
+   reconciles the work package and returns versioned external-reference and
+   retry/conflict evidence. Neither consumer writes connector tables.
+
+Every executable collaboration connector must pass the common connector
+contract checks plus a provider-focused minimum proof:
+
+- optional-module startup and partial compositions work without the provider;
+- profile health uses secret references and redacts credentials and remote
+  response bodies;
+- tenant/resource authorization is checked before discovery, provisioning,
+  lookup, synchronization, or evidence capture;
+- dry-run/simulation performs no remote mutation and explains unsupported
+  operations;
+- create/publish calls are idempotent, retries preserve the same external
+  reference, and outcome-unknown or version conflicts remain reconcilable;
+- callbacks/webhooks verify authenticity, tenant/profile binding, replay
+  protection, and bounded payloads;
+- disable/retire behavior revokes new use while preserving non-secret audit and
+  external-reference evidence;
+- Collabora/OnlyOffice prove discovery plus one non-production editing-session
+  round trip; Matrix/Mattermost/Rocket.Chat prove room lookup/create plus one
+  authenticated bounded event; Jitsi/BigBlueButton prove conference
+  provision/cancel; OpenProject proves project/work-package lookup, idempotent
+  publish, and conflict handling.
+
+These are connector acceptance tests, not a claim that those connectors are
+already implemented. Their implementation state remains in the owning
+connector issues and catalogue.
+
 ### Payment And Public Cashier Systems
 
 - Strategy: integrate/export/import; keep the payment provider or cashier as

docs/RECORDS_FILING_CONTRACT.md

@@ -0,0 +1,94 @@
+# Records Filing Contract
+
+Core exposes a small provider-neutral contract for filing exact source
+revisions into an institutional record. Core does not own records semantics,
+source-object authorization, or source bytes. `govoplan-records` owns filing
+orchestration and chronology; each source module owns resolution of its exact
+revision.
+
+## Capability Names
+
+- `records.filing` is supplied by the enabled Records module.
+- `records.source.<module>` is supplied by an enabled source module, for
+  example `records.source.files` or `records.source.cases`.
+- `records.archive.<provider>` is supplied by an enabled archive-transfer
+  adapter. Discovery does not imply conformance or current health.
+
+Callers discover capabilities through the module registry. They must not
+import optional source-module internals.
+
+## Exact Source Identity
+
+`RecordSourceLocator` identifies one tenant, source module, resource type,
+resource ID, and immutable source revision. A source provider must:
+
+1. reject cross-tenant resolution;
+2. require a non-empty purpose;
+3. re-evaluate the caller's current module and object authorization;
+4. resolve exactly the requested revision, never a mutable "current" alias;
+5. return safe display/provenance metadata and a SHA-256 digest when the source
+   has stable bytes or a canonical snapshot;
+6. fail closed when the revision is missing, quarantined, corrupt, or no longer
+   authorized.
+
+Historical Records browsing never revives historical access rights. The
+source's current authorization decision remains authoritative when filing.
+
+## Filing Semantics
+
+`RecordFilingRequest` binds the exact source to a record, purpose, filing
+reason, relationship, institutional context, and idempotency key. Records must
+persist source identity and resolution evidence together with the filing actor,
+represented capacity, valid time, recorded time, and immutable chronology.
+
+An idempotency key may replay only an identical request. A conflicting reuse
+must fail. Filing does not transfer ownership of source content and must not
+silently copy mutable source state.
+
+## Versioning
+
+The Python DTOs and protocols live in `govoplan_core.core.records`. The
+manifest interface `records.filing` starts at `1.0.0`. Incompatible DTO or
+behavior changes require a new interface version and release impact analysis;
+additional optional metadata remains backward compatible.
+
+## Initial Providers
+
+- Files resolves an exact managed `FileVersion`, verifies current Files access
+  and blob integrity, and returns its stored content digest.
+- Cases resolves an exact immutable case revision after current case access and
+  returns a digest of the canonical revision snapshot.
+
+Provider-specific selection UI belongs to the source module. The generic
+Records dialog remains a diagnostic/manual fallback for exact identifiers.
+
+## Archive Transfer Boundary
+
+`RecordTransferPackage` binds a stable package ID, record revision, provider
+profile, canonical manifest, and manifest SHA-256. An archive provider exposes
+`RecordArchiveProviderState` before dispatch and accepts only a
+`RecordArchiveTransferRequest` for a declared healthy profile. Its receipt must
+identify the same package and provider and return one bounded outcome:
+`accepted`, `rejected`, or `outcome_unknown`.
+
+An unknown outcome is never retry-safe. Callers must retain the intent and
+reconcile it against the provider before another effect. Provider state also
+declares authority mode, freshness, limitations, and whether the provider is a
+simulation. Credentials, transport configuration, archive-specific package
+schemas, and custody semantics remain provider-owned.
+
+Records includes `records.archive.simulation` to prove package and receipt
+handling. The simulation is explicitly non-conformant, transfers no custody,
+and cannot be used as evidence of an archive handoff. A real provider requires
+a selected target/profile, provider-specific recovery declaration, and target
+test evidence.
+
+## Form Evidence Boundary
+
+Form attachments use the separate provider-neutral contract in
+`govoplan_core.core.form_evidence`. Forms Runtime requests short-lived,
+purpose-bound upload grants and re-inspects the exact provider-owned evidence
+before final submission. The provider keeps byte storage, quarantine,
+classification, and retention ownership; Forms Runtime stores only immutable
+evidence references and bounded verification results. This contract is not an
+alternative path for Records filing or archive custody.

docs/RELEASE_DEPENDENCIES.md

@@ -95,6 +95,15 @@ package entry. Keep one committed full-product release lockfile at
 `webui/package.release.json` in a clean release workspace. Development
 `package-lock.json` may continue to point at local `file:` dependencies.
 
+The default WebUI build discovers every declared, installed module package,
+including Tasks and its Work page, dashboard widget, and Quick Access tool.
+Discovery still respects enabled backend modules and user permissions; adding
+a package never grants access. `GOVOPLAN_WEBUI_MODULE_PACKAGES` selects an
+explicit smaller build when set, and an explicitly empty value selects
+core-only. The prebuild interface check compares the default descriptor list
+with the package manifest so explicit permutation tests cannot conceal a
+module accidentally omitted from the ordinary build.
+
 Frontend module permutations are regression-tested through
 `GOVOPLAN_WEBUI_MODULE_PACKAGES` and temporary build output, not through
 committed lockfiles for every possible combination. If a smaller composition
@@ -197,6 +206,13 @@ If both file and URL are set, the URL wins. The cache is used when a remote
 fetch fails, so an operator can still inspect the last known catalog. A cached
 catalog must still pass signature, freshness, channel, and replay validation.
 
+If neither source is configured, the Admin package directory discovers the
+official public stable catalog at
+`https://govoplan.add-ideas.de/catalogs/v1/channels/stable.json`. Core verifies
+that fallback against the public key pinned in the installed Core package. An
+explicit deployment catalog always takes precedence; a configured source that
+is unavailable or invalid fails closed instead of silently falling back.
+
 An official catalog is a JSON object with:
 
 - `catalog_version`
@@ -212,6 +228,14 @@ Each module entry can declare:
 
 - backend package name and pinned install reference
 - WebUI package name and pinned install reference
+- `artifact_integrity` for each package, including the HTTPS registry URL,
+  filename, byte size, SHA-256, package identity, source tag, and source commit
+- `source`, binding the repository and immutable tag/commit identity, with
+  optional HTTPS repository and revision links
+- `availability`, either `available` or `withdrawn`; a withdrawn entry must
+  carry an operator-readable `availability_reason` and cannot be planned
+- `configuration_requirements` and an optional HTTPS `release_notes_url` for
+  prerequisites and release-specific operator guidance
 - display metadata and tags
 - `license_features`, the feature entitlements required to plan that install
 - `dependencies` and `optional_dependencies`, the module ids expected in the
@@ -239,6 +263,12 @@ Each module entry can declare:
 - `requires_interfaces`, named interface contracts and version ranges required
   by this module
 
+Core validates these fields before exposing the directory. Admin derives a
+read-only catalog state from the installed package set, catalog dependency
+closure, named-interface providers, current-version window, availability, and
+generic license policy. This is an early operator diagnostic; trusted installer
+preflight remains the authoritative mutation gate.
+
 The signature is Ed25519 over canonical JSON with both `signature` and
 `signatures` removed. Core accepts the legacy single `signature` field and the
 new `signatures` array.
@@ -302,6 +332,12 @@ Catalog provenance changes preflight severity:
   plans, so operators can still use offline or emergency package refs
 - valid-catalog warnings, such as intentionally unsigned local catalogs when
   signature enforcement is disabled, remain warnings
+- a saved catalog plan must match the currently validated entry exactly;
+  altered package refs, artifact identities, channel, sequence, trust state, or
+  signing-key identity block the run and require replanning
+- a trusted remote artifact is downloaded before mutation into a private
+  SHA-256-addressed installer cache, checked for exact size and digest, and
+  passed to `pip` or npm only as that verified local file
 - selected catalog entries with unsatisfied non-optional named interface ranges
   block activation before the installer runs
 - selected catalog entries whose target dependencies are neither installed nor
@@ -519,6 +555,11 @@ Catalog entries can require license features:
 Core checks those requirements against an offline license file before allowing
 the entry into the install plan.
 
+Official open-source GovOPlaN entries do not declare license features. The
+license contract remains generic for external catalogs, deployment presets,
+configuration/package directories, and support offerings; it gates only an
+entry that explicitly asks for a feature.
+
 ```bash
 GOVOPLAN_LICENSE_FILE=/srv/govoplan/license.json
 GOVOPLAN_LICENSE_ENFORCEMENT=true

docs/SECURITY_AUDIT.md

@@ -12,4 +12,6 @@ tools/checks/security-audit/run.sh --mode full --scope govoplan
 
 Canonical documentation:
 
-- `/mnt/DATA/git/govoplan/docs/SECURITY_AUDIT.md`
+- `/mnt/DATA/git/govoplan/docs/operations/SECURITY_AUDIT.md`
+
+Implementation contract: [disposable resource-bounded operations](BOUNDED_PROCESS_CONTRACT.md).

docs/SEMANTIC_DOCUMENTATION_SUBJECTS.md

@@ -0,0 +1,118 @@
+# Semantic Documentation Subjects
+
+## Purpose And Ownership
+
+The semantic-documentation subject contract lets an optional module expose the
+configured artifacts that administrators may document: for example a form, a
+form field, a workflow, or a workflow state. It is a discovery and resolution
+contract, not a second configuration API.
+
+The module that owns an artifact also owns its subject provider, authorization,
+identity, revision, route, and lifecycle semantics. Docs may discover those
+providers through Core and attach authored documentation to their stable
+references. Docs must not import the feature module, read its tables, or copy
+configuration content into a generic index.
+
+This contract is additive to manifest `DocumentationTopic` contributions and
+configured-state `documentation_providers`. Every providing module must retain
+static user and administrator documentation baselines. The baselines explain
+the feature even when the provider is disabled, unavailable, or has no
+configured subjects.
+
+## Identity And Versioning
+
+`SemanticDocumentationSubjectReference` identifies a subject with:
+
+- owning module and tenant;
+- a module-defined subject kind and stable identifier;
+- an optional typed nested anchor, such as `field/registration-number`;
+- the revision and canonical fingerprint observed when documentation was
+  authored or reviewed.
+
+The `stable_key` derives only from identity. A rename or configuration revision
+therefore does not detach existing documentation. A nested anchor has its own
+identity so a field can be documented independently from its form.
+
+Providers must resolve an old reference as one of:
+
+- `available`: the observed revision/fingerprint is still current;
+- `changed`: the same stable subject has changed and may need review;
+- `superseded`: another stable reference replaced it;
+- `missing`: the subject was removed or is no longer resolvable;
+- `temporarily_unavailable`: the provider cannot currently determine state.
+
+Absence is not authorization. A provider returns `None` when the principal may
+not learn whether a subject exists. Core also rejects cross-tenant list and
+resolution requests before calling a provider.
+
+## Safe Projection
+
+Descriptors contain only bounded, explicit presentation fields: localized
+labels and descriptions, breadcrumbs, a local route, audience,
+classification, and required scopes. They must not contain credentials,
+personal data, arbitrary provider metadata, configuration payloads, or the
+authored documentation itself. Routes are application-local and are still
+subject to normal route authorization.
+
+The fingerprint is a review signal, not a concurrency token or a content hash
+that callers may use to reconstruct configuration. Providers should calculate
+it from the smallest canonical JSON projection whose semantic changes require
+documentation review. Volatile timestamps and secrets must be excluded.
+
+## Provider Registration
+
+A provider is registered under its exact module-scoped capability name:
+
+```python
+from govoplan_core.core.modules import CapabilityDocumentation
+from govoplan_core.core.semantic_documentation import (
+    SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION,
+    semantic_documentation_subject_capability,
+)
+
+capability = semantic_documentation_subject_capability("forms")
+
+manifest = ModuleManifest(
+    id="forms",
+    # ...
+    capability_factories={capability: build_semantic_subject_provider},
+    capability_documentation={
+        capability: CapabilityDocumentation(
+            label="Form semantic subjects",
+            summary="Lists authorized configured forms and fields for Docs.",
+            contract_version=SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION,
+            documentation_types=("admin", "user"),
+        )
+    },
+    documentation=(admin_baseline, user_baseline),
+)
+```
+
+The capability is `documentation.semantic_subjects.<module_id>`. Registry
+validation rejects a mismatched owner, missing capability documentation, a
+wrong contract version, or missing static baselines.
+
+`list_semantic_documentation_subjects` performs authorized, paginated discovery
+across installed providers. `resolve_semantic_documentation_subject` targets
+one owner without loading another feature module. Providers must apply the
+current tenant and principal on every call and must not infer visibility from a
+previous list result.
+
+## Lifecycle And Integration Rules
+
+- Keep subject and anchor identifiers stable across display-name and route
+  changes.
+- Return `superseded` only with the replacement reference; do not silently
+  rewrite stored references.
+- Return a reason code for missing or temporarily unavailable subjects without
+  exposing sensitive detail.
+- Reauthorize both discovery and resolution. Stored documentation references
+  confer no access to a live artifact.
+- Treat a changed fingerprint as a request for editorial review. It does not
+  automatically invalidate or publish authored documentation.
+- Removing a feature module leaves references resolvable as provider
+  unavailable. Docs can preserve history without importing the module.
+
+Forms, Workflow, and later modules should implement their subject providers in
+their own repositories. Docs owns the authored semantic-documentation records,
+review workflow, and projection UI.

docs/STATE_AND_RECOVERY_CONTRACT.md

@@ -93,10 +93,17 @@ Alembic, and post-migration tasks. The lock is session-scoped and therefore
 released if the migration process dies.
 
 Runtime roles run `govoplan_core.commands.wait_for_database`. It waits until
-the database has exactly the configured Core/module Alembic heads and never
-upgrades schema. This permits a migration Job and runtime Deployments to be
+the database has exactly the configured, dependency-resolved Core/module
+Alembic heads and never upgrades schema. Cross-module `depends_on` revisions
+therefore do not leave runtime roles waiting for a branch marker Alembic has
+correctly consumed. This permits a migration Job and runtime Deployments to be
 submitted together while keeping startup fail-closed.
 
+Runtime coordination records the installed `govoplan-core` distribution
+version for API, worker and scheduler roles. FastAPI/OpenAPI metadata versions
+are presentation metadata and must not be used as deployable software identity;
+mixing the two would create a false version-skew readiness failure.
+
 ## Recovery Ledger
 
 `govoplan_core.core.recovery` provides a durable operation and evidence
@@ -133,7 +140,7 @@ Core does not create production database backups. The deployment owner must
 provide backup, retention, encryption, restore verification, and recovery-point
 coordination for PostgreSQL, object storage, and encryption keys. The canonical
 operator procedure is documented in
-`govoplan/docs/RECOVERY_AND_ROLLBACK_GUARANTEES.md`.
+`govoplan/docs/operations/RECOVERY_AND_ROLLBACK_GUARANTEES.md`.
 
 ## Verification
 

docs/TABULAR_SOURCE_CONTRACT.md

@@ -19,3 +19,72 @@ Connector health and preview diagnostics must contain no credentials, endpoint
 userinfo, row values, or unbounded remote error bodies. A Datasource origin
 preserves this contract so registration and staging do not erase source mode,
 health, pushdown, or preview-limit evidence.
+
+## Durable CSV imports and original evidence
+
+`TabularCsvSource` optionally accompanies a durable `TabularSnapshotInput` or
+`DatasourceStageInput`. It carries the exact submitted Unicode text, delimiter,
+explicit value mode and parser profile. It is not part of ordinary catalogue,
+preview or stage DTOs. Transient inspection does not retain original content.
+
+The `text` mode preserves cell strings, including whitespace, leading zeroes,
+decimal spelling, boolean-looking text and explicit empty cells. It rejects
+malformed quoting and rows with missing or extra cells. Header normalization is
+unchanged. The API default remains `legacy_typed` for existing integrations;
+interactive CSV imports offer text mode by default and an explicit legacy choice.
+JSON and existing stored snapshots are not reinterpreted. Core and Datasources
+retain distinct versioned legacy parser profiles where their historical coercion
+rules differ. Shared schema inference preserves first-seen column order, missing
+value nullability and the owning provider's type naming.
+
+Owners verify that the source reparses to exactly the stored projection, including
+scalar types: `true`, `1` and `1.0` are not equivalent evidence. Raw input and row
+projections each have a 5,000,000-byte limit; row parsing is capped at 10,000 rows.
+The original text has its own UTF-8 SHA-256 and byte count, separate from the
+existing row fingerprint. Only a small allowlisted source summary enters metadata.
+Checksums detect drift; they are not digital signatures or protection against an
+attacker who can rewrite the database and all its evidence.
+
+Original exports are explicit owner APIs, tenant scoped, integrity checked and
+`no-store`. Datasources additionally requires administrator scope, audits the
+export, and denies the whole original when current or historical governance
+restricts any row or field. Freezing verifies the prior summary before copying
+source evidence and retains prior policy restrictions, including referenced
+policy evaluations. At most 32 distinct governance snapshots may accompany one
+original; further incompatible history fails explicitly. Payload disposal also
+disposes retained original content. Connectors applies its own current read,
+tenant and lifecycle checks. See the owning module's documentation for endpoints.
+
+Original UTF-8 text is not proof of pre-decoding file bytes, and does not undo
+CSV spreadsheet formula semantics. Exported content is deliberately unmodified;
+operators must treat it as untrusted input when opening it in a spreadsheet.
+New nullable columns require the owning modules' additive migrations. Historical
+rows remain unchanged and report original content unavailable, not reconstructed.
+Back up retained originals before any schema downgrade that removes those columns.
+
+## Deutsch: CSV-Datentreue
+
+Dauerhafte CSV-Importe können den unveränderten übermittelten Unicode-Text mit
+Trennzeichen, Parserprofil und explizitem Wertemodus aufbewahren. Der Textmodus
+erhält Zellwerte einschließlich Leerzeichen, führender Nullen und Dezimalschreibweise.
+Fehlerhafte Zeilen werden abgewiesen. Die API bleibt aus Kompatibilitätsgründen bei
+der bisherigen Typumwandlung als Standard; im Importdialog ist Text voreingestellt.
+Bestehende Daten und JSON-Importe werden nicht neu interpretiert.
+
+Original und Zeilenprojektion werden getrennt begrenzt und geprüft; boolesche
+Werte, Ganzzahlen und Gleitkommazahlen sind keine austauschbaren Belege. Der
+Originaltext erscheint weder im Katalog noch in Vorschauantworten. Die expliziten
+Export-APIs prüfen Mandant, Berechtigungen, Lebenszyklus und gespeicherten Hash.
+Datasources verlangt zusätzlich Administrationsrechte, protokolliert Exporte und
+berücksichtigt aktuelle sowie historische Zeilen-, Feld- und Zugriffsrichtlinien.
+Eingeschränkte Originale werden vollständig gesperrt, nicht teilweise freigegeben.
+Eingefrorene Kopien übernehmen diese Einschränkungen; nach 32 unterschiedlichen
+Richtlinienständen wird eine weitere Kopie mit neuer Richtlinie explizit abgewiesen.
+Die Aufbewahrungsbereinigung entfernt auch das gespeicherte Original.
+
+Die Grenzen betragen jeweils 5.000.000 UTF-8-Bytes für Original und Projektion
+sowie 10.000 Zeilen. Prüfsummen sind keine Signaturen. Ein CSV-Original bleibt beim
+Export unverändert und kann Tabellenkalkulationsformeln enthalten. Frühere
+Dateikodierungen lassen sich daraus nicht rekonstruieren. Additive Migrationen
+ändern keine historischen Zeilen; fehlende Originale werden nicht erfunden.
+Vor einem Schema-Downgrade sind aufbewahrte Originale zu sichern.

docs/TEMPORAL_DATA_CONTEXT.md

@@ -0,0 +1,68 @@
+# Temporal Data Context
+
+GovOPlaN exposes one read context for data validity and system knowledge. The
+calendar control in the authenticated titlebar applies that context to
+supported list and detail reads for the current account and tenant.
+
+## Two Independent Axes
+
+- **Valid time** answers when a fact applied in the represented domain.
+- **Recorded time** answers what the system had recorded by a particular
+  instant.
+
+The default is data valid now under the latest recorded state. `At time`
+selects a valid-time instant. `All` removes the valid-time interval filter but
+still uses the selected recorded state. The optional recorded-state cutoff can
+be combined with any valid-time mode, which keeps correction history distinct
+from changes in real-world validity.
+
+An interval is half open: `valid_from <= instant < valid_to`. A revision belongs
+to a recorded-state snapshot when `recorded_at <= cutoff` and it was not
+superseded at or before that cutoff.
+
+## Security And Mutation Rules
+
+The temporal data context is a read projection, not an authorization context.
+Authentication, permissions, active delegations, tenant boundaries, module
+policy, and maintenance controls are always evaluated under current security
+state. A historical projection never restores an expired permission.
+
+The context also does not supply mutation dates. Writes continue to target the
+current lifecycle revision and must carry their explicit valid/effective dates,
+expected revision, reason, and evidence where the owning contract requires
+them. A screen showing historical data must not silently turn a normal edit
+into a historical correction.
+
+## HTTP Contract
+
+Core accepts these request headers:
+
+| Header | Meaning |
+| --- | --- |
+| `X-Govoplan-Validity-Mode` | `current`, `at`, or `all` |
+| `X-Govoplan-Valid-At` | Timezone-aware ISO 8601 instant required by `at` |
+| `X-Govoplan-Recorded-At` | Optional timezone-aware system-knowledge cutoff |
+
+Invalid or naive timestamps fail with HTTP 400. Responses expose the resolved
+mode and evaluated instant. Conditional JSON responses vary by all three
+request headers, and the shared WebUI API client includes them in request
+deduplication and conditional-cache keys.
+
+## Module Adoption
+
+Revision-owning modules apply
+`govoplan_core.db.temporal.apply_temporal_revision_filter` only to read queries
+that are meant to follow the platform context. Explicit version references and
+explicit resolver `effective_at` arguments take precedence. Current-row
+lookups used for optimistic concurrency, authorization, routing, effects, or
+other mutations must remain explicit and context-independent.
+
+The initial bitemporal adoption covers Decisions, Mandates, Parties, and
+Services. Their immutable revisions have indexed valid, recorded, and
+superseded timestamps. Modules with effective-dated security records or
+recorded-only revision histories require separate display-query adoption so
+the global selector cannot affect current authorization or execution.
+
+The WebUI selection is stored in session storage per account and tenant. A
+change remounts the active module route so existing page loaders issue a fresh
+request. Returning both axes to their defaults removes the stored selection.

docs/THEMING.md

@@ -4,6 +4,9 @@ GovOPlaN supports `system`, `light`, and `dark` as persisted user preferences.
 `system` follows `prefers-color-scheme` live; it is not resolved permanently at
 save time. Core applies the resolved mode through `data-theme` on the document
 root and exposes the selected preference through `data-theme-preference`.
+Each user may also choose a validated `default`, `civic_blue`, `forest`, or
+`plum` accent palette. Core applies it through `data-palette`; every module
+inherits the result through semantic tokens without module-specific CSS.
 
 ## Ownership
 
@@ -12,17 +15,48 @@ root and exposes the selected preference through `data-theme-preference`.
 - Modules consume semantic tokens such as `--surface`, `--text`, `--line`, and
   the status token families. They may define domain aliases whose values resolve
   to shared tokens.
-- User preference selects the mode. Tenant and system policy may provide a
-  future default, but must not silently replace an explicit user choice.
-- Tenant branding is a separate policy surface and must preserve contrast and
-  status semantics in both modes.
+- Palette defaults form a provenance chain: system, tenant, then an explicit
+  user choice. Invalid stored values are ignored. Reset means inheritance and
+  does not copy the current parent value into the child scope.
+- A policy lock is separate from the default. A system lock wins over every
+  child scope; otherwise a tenant lock suppresses a personal override. The
+  authenticated profile reports the effective palette, source, inherited
+  palette, and lock state.
+- Advanced personal overrides are a separately governed surface. The system
+  must opt in, a tenant may inherit or block that decision, and palette locks
+  always suppress overrides. Changing either policy requires
+  `admin:policies:write` in addition to the owning settings permission.
+
+## Palette safety and scope
+
+The Settings preview shows the chosen or inherited accent in every applicable
+light/dark preview before Save. Presets are checked for WCAG AA contrast in the
+theme contract. When policy permits, the shared advanced editor can atomically
+override accent, surface, and semantic status pairs for both modes. Every
+foreground/background pair must meet WCAG AA contrast, and success,
+information, warning, and danger colors must remain distinct. Invalid stored
+documents fail closed and are not partially applied.
+
+Import and export use the exact versioned JSON schema `schema_version: "1"`.
+Both `light` and `dark` must contain every supported token exactly once as a
+six-digit hex value. Import changes only the local draft; Save persists the
+whole document. Removing overrides returns to palette and policy inheritance.
+The system default is disabled so upgrades do not unexpectedly admit arbitrary
+branding. Tenant `null` means inherit, `false` blocks, and `true` is accepted
+only while the system permits overrides.
 
 Do not introduce fixed foreground/background colors in a module merely to make
 one mode look correct. Add or reuse a semantic Core token, then define both
 light and dark values. Bitmap content and externally authored HTML are exempt,
 but their surrounding controls must still use the shared tokens.
 
-`npm run test:theme-contract` verifies the root behavior and representative
+`npm run test:theme-contract` verifies root mode/palette behavior, preset and
+custom-override validation/application, and representative
 Campaign, Calendar, Files, and Mail token consumption. The check runs before a
 production WebUI build.
 
+Runtime validation and token application live in the dependency-free
+`webui/src/components/appearanceOverrides.ts`; both the shell and the shared
+editor use it. The shell must not import the editor to apply an existing theme:
+settings controls load with their route, while valid saved colors apply
+synchronously and invalid documents still fail closed before any token is set.

docs/TICKET_INTEGRATION_CONTRACTS.md

@@ -0,0 +1,59 @@
+# Ticket Integration Capability Contracts
+
+Core owns two narrow, optional contracts that let the Tickets module compose
+with policy and formal-procedure modules without importing either one. Tickets
+remains the authority for operational ticket identity, lifecycle, assignment,
+comments, links, and immutable history.
+
+## Capability Names
+
+- `tickets.routing` optionally supplies a `TicketRoutingProvider`.
+- `tickets.case_escalation` optionally supplies a
+  `TicketCaseEscalationProvider`.
+
+Both contracts are version 1 and are defined in
+`govoplan_core.core.tickets`. Registry helpers return `None` when a capability
+is absent or has the wrong shape, so optional-module absence is normal runtime
+state rather than a startup failure.
+
+## Routing
+
+Tickets sends a bounded, tenant-scoped `TicketRoutingRequest` containing the
+ticket reference, type, priority, title, receive time, optional queue hint, and
+non-secret attributes. The provider returns its identity and may return a queue
+reference, timezone-aware service target, human-readable explanation, and
+bounded metadata.
+
+The provider is advisory. Tickets snapshots any returned queue and target into
+its own record and history. An absent provider, a no-match plan, or an absent
+queue must not prevent ticket intake; authorized staff can route manually.
+Providers must not persist a second ticket lifecycle.
+
+## Case Escalation
+
+Tickets sends a `TicketCaseEscalationCommand` with stable tenant, ticket, and
+display references, the requested Case type, actor-visible handoff note,
+timezone-aware occurrence time, and an idempotency key. The provider returns a
+stable Case identifier, number, bounded application-relative URL, replay flag,
+and bounded metadata.
+
+Providers must:
+
+- recheck tenant and Case-creation authorization;
+- reject an absent or inactive requested Case type;
+- make identical retries resolve the same Case;
+- preserve the Ticket reference in governed Case context; and
+- return only an application-relative path, never an untrusted external URL.
+
+Tickets records the result and its own escalation evidence. Cases remains the
+authority for the formal procedure; Tickets remains the authority for the
+operational request. Creating a Case does not merge or silently close either
+lifecycle.
+
+## Failure And Transaction Semantics
+
+Capability calls receive the caller's active persistence session so a concrete
+provider can participate in the same unit of work. Authorization and validation
+errors fail the requested routing/escalation mutation explicitly. The caller
+must still apply its own permission checks, tenant boundary, replay protection,
+and immutable evidence rules.

docs/UI_DESIGN_PRINCIPLES.md

@@ -0,0 +1,150 @@
+# Product UI design principles
+
+These rules govern every GovOPlaN module, including administration, tenant and
+user settings, public forms, widgets, dialogs and shared shell surfaces. Core
+owns the reusable components; modules own their meaning and authorized behavior.
+They complement the [page composition contract](PAGE_LAYOUT_USAGE_GUIDELINES.md).
+
+Initial rule revision: **UI-2026-09-08**. Record this identifier in review evidence;
+subsequent rule changes need a dated change reference and a UI-09 propagation check.
+
+## UI-01 — Help belongs to visible text
+
+Place a small documentation book immediately after the heading or label whose
+meaning it explains. For example, the Dashboard book belongs to **Dashboard**,
+not the Reload, Edit or Save group. The same applies to page, pane, card and
+dialog titles, field labels, widget summaries and contextual explanations.
+Do not put an isolated documentation icon in an action bar or in an otherwise
+empty row. A navigation icon is not a contextual documentation link.
+
+Use `titleHelp={<DocumentationHelpLink reference={...} />}` on `PageLayout`,
+`PageHeader`, `PageTitle`, `AdminPageLayout`, `Card` or `Dialog`. Full-canvas
+workspaces use the visible `title` and `titleHelp` of `WorkspaceActionBar` (or
+`PageActionBar`), separate from its action groups. Use `TextWithHelp` around
+existing text for smaller contexts; fields use `FormField`/`FieldLabel` and
+their documentation reference. Do not duplicate the heading to add help.
+
+Dashboard widgets provide their owning `documentation` reference in
+`DashboardWidgetContribution`; Dashboard renders it beside the existing widget
+card title. Do not add a second widget heading or a footer-only documentation icon.
+
+Keep the icon outside the heading's accessible name and outside other buttons
+or links. It needs an accessible purpose, keyboard focus and a usable hit area.
+Long German headings must wrap without colliding with actions. Collapsing a
+card or configuring a dashboard must not change which title the book explains.
+The existing optional Docs-module/hosted-documentation fallback remains intact;
+placement does not change permissions or require the Docs module.
+
+## UI-02 — Display first; edit deliberately
+
+Ordinary overview and detail pages show compact, readable data. Group related
+facts with shared description, summary and card components. Offer a clear Edit
+action that opens a scoped dialog for a coherent group of settings. Avoid
+presenting every possible input, select and save control simultaneously.
+
+The Campaign dashboard is a priority application: show the effective settings
+and their state compactly, then edit related settings in focused dialogs. This
+is review work to implement, not a claim that all current screens already comply.
+
+An explicit editing mode is a justified exception for genuinely broad editing,
+large tables, graphical designers or other tasks poorly served by a modal.
+Record why the exception helps the user. Keep reading and editing distinguishable;
+enter editing deliberately and preserve shared Save/Cancel, validation, conflict,
+permission and unsaved-draft behavior. Filters, search and selection do not need
+an edit dialog: they change the view rather than the underlying data.
+
+Cancel/Close must work even without changes. Never save a draft merely because
+a dialog opens, selection changes or navigation happens. Commit only the intended
+fields; do not let unrelated incomplete settings block an independent edit.
+
+## UI-03 — Predictable actions and consequences
+
+Follow the shared semantic page/pane action contract: context on the left;
+Reload on refreshable surfaces followed by Create on the right. Editable drafts
+have the standard state, Cancel/Discard and Save controls. Keep positions stable
+across selection and loading changes, and explain unavailable applicable actions.
+Separate destructive actions visually and semantically. Reload is a read, not
+an implicit synchronization, import, send or other mutation.
+
+## UI-04 — Shared geometry and readable collections
+
+Use central page, card, table, form-grid and dialog primitives. Table cards use
+`bodyLayout="table"`, spanning the card body without negative margins or expanded
+`calc()` widths. Keep row actions reachable and deliberate column resizing
+predictable. Avoid redundant taglines, arbitrary gaps, fixed dialog widths and
+horizontal scrolling where content can wrap. Use actual pagination and shared
+list filters; every result and every validation detail must remain reachable.
+
+## UI-05 — Stable feedback and loading
+
+Preserve useful loaded content on refresh failure; distinguish initial loading,
+empty, unavailable, stale, blocked, failed and successful states. Use the shared
+scoped loading overlay. Long operations show truthful processed/total and outcome
+counts when known, not simulated progress or whole-page background reloads.
+Prevent duplicate submissions without trapping the user in a clean editing mode.
+
+## UI-06 — Predictable navigation and selection
+
+Use shared breadcrumbs, side rail and tree contracts. Tree disclosure controls
+expand/collapse; labels select. Track the clicked occurrence when one topic or
+record appears in several branches. Preserve context, filters and selection
+where appropriate, without carrying data across authorization or tenant scopes.
+
+## UI-07 — Accessibility and German parity
+
+Check keyboard order, visible focus, accessible names, dialog focus restoration,
+contrast, zoom and narrow layouts. Status must not rely on color alone. Use
+plain, task-oriented wording and complete German UI and module-owned help for
+the changed workflow. Long labels must not hide actions or force needless scroll.
+
+## UI-08 — Integrity and permissions are not visual shortcuts
+
+Preserve authorization, versioning, audit evidence, immutable historical data and
+safe retries. Do not remove a guard to make an action look available. Keep drafts
+recoverable after failures and conflicts. Loading, cancellation and dialog
+transitions must not silently write, discard or overwrite data. Never verify UI
+changes by sending real mail or changing production records without authorization.
+
+## UI-09 — Review incrementally and propagate lessons
+
+The [Meta UI-review epic](https://git.add-ideas.de/GovOPlaN/govoplan/issues/56)
+links one review ledger per module. Each starts pending, including modules
+touched by a focused shared-component pass. An implemented help-placement rule
+does not constitute a full module review.
+
+Inventory the module's surfaces, inspect them with realistic data and permissions,
+record findings and decisions, implement bounded fixes, then attach automated
+and manual evidence. Distinguish **to do**, **implemented**, **verified**, and
+**accepted exception**. Record the applicable principle IDs and actual screen
+coverage; do not mark a module complete from a code search alone.
+
+When a review reveals a reusable lesson, update this document and the central
+component/check first where possible. Record the rule change in Meta and create
+a follow-up or reopen affected already-reviewed modules. The review ledger must
+show which rule revision each module has actually verified. Prefer finishing
+the active review slice before expanding functionality, while recording unrelated
+bugs and decision-dependent work without losing them.
+
+## Kurzfassung auf Deutsch
+
+- **UI-01:** Das Dokumentationsbuch steht direkt rechts neben der zugehörigen
+  Überschrift oder Beschriftung, nicht zwischen Aktionsschaltflächen.
+- **UI-02:** Daten zunächst kompakt anzeigen; zusammengehörige Einstellungen
+  gezielt in einem Dialog bearbeiten. Ein ausdrücklich aktivierter
+  Bearbeitungsmodus, etwa für große Tabellen, ist eine begründete Ausnahme.
+  Abbrechen funktioniert auch ohne Änderungen; Änderungen werden bewusst gespeichert.
+- **UI-03–UI-08:** Einheitliche Aktionen und Abstände, erreichbare Tabelleninhalte,
+  verständlicher Fortschritt, vorhersehbare Navigation, Barrierefreiheit und
+  deutsche Texte dürfen Berechtigungen oder Datenintegrität nicht schwächen.
+- **UI-09:** Alle Module werden einzeln geprüft. Neue Gestaltungsregeln werden
+  auch in bereits geprüften Modulen nachgezogen und mit Nachweisen dokumentiert.
+
+## Automated coverage and its limits
+
+`tools/checks/check-heading-help.mjs` in Meta checks JSX documentation-link
+placement throughout the workspace, including aliases and simple local variables.
+Core component and browser tests check heading association, accessible names,
+focus, collapse/loading transitions and narrow/wide layouts. Existing layout,
+action, dialog and DataGrid checks continue to apply. These tests protect shared
+contracts; they do not certify every module's usability or replace the individual
+screen reviews and their documented exceptions.

docs/UI_UX_DECISION_LEDGER.md

@@ -51,10 +51,14 @@ contestability, responsibility, and traceability at the point of action.
 | UX-025 | `window.alert` and the global `alert` function are prohibited. A narrowly necessary exception requires product-owner authorization and an entry in the alert exception register before implementation. | Accepted | All WebUI code |
 | UX-026 | A table defines one stable ordered action set. A row-level unavailable action remains in its normal position and is disabled, preferably with `disabledReason`; structurally irrelevant actions are omitted for the entire table. Empty rows reserve the same slots so their Add action stays in the normal left-most action position. | Accepted | All structured tables |
 | UX-027 | The platform icon rail keeps its brand header and utility footer visible. Only the module-navigation region scrolls when installed and permitted modules exceed the available viewport height. | Accepted | Core WebUI shell |
-| UX-028 | Maintenance and offline state change the titlebar surface and repeat a quiet status label behind its controls. They must not replace, cover, or intercept the centered global-search surface; an accessible status control remains in the leading titlebar area. | Accepted | Core WebUI shell |
+| UX-028 | Maintenance and offline states use their established textual warning banners centered in the titlebar. They must not recolor the shell or add decorative status icons. Global search is a right-side command, so warnings do not replace its trigger or overlay. | Accepted | Core WebUI shell |
 | UX-029 | Recoverable page and module errors use the central compact `DismissibleAlert` presentation with an explicit recovery action where one exists. Full-height workspaces must overlay page feedback instead of allowing an alert to become a stretched workspace row. | Accepted | Core and module WebUIs |
 | UX-030 | At narrow widths, the titlebar uses separate context and command rows. Context selectors remain horizontally reachable, search retains its compact trigger, and language/help/notification/account commands remain fixed icon controls without overlap. Shared content padding contracts so domain workspaces retain usable width. | Accepted | Core WebUI shell and all module workspaces |
 | UX-031 | Public controls and extension contributions use stable, module-namespaced interface identities. Shared controls expose `interfaceId` and `helpTopicId`; generated source anchors are inventory evidence, not a substitute for an explicit ID when documentation, policy, or automation refers to the control. | Accepted | Core and module WebUIs |
+| UX-032 | `F1` resolves help from the focused field or action, then its dialog/section/page and registered route. Focused contexts retain the page fallback; Docs applies audience and permission filtering and falls back to visible module documentation. | Accepted | Core shell, Docs, and all module WebUIs |
+| UX-033 | Global search is the left-most titlebar command, immediately before language selection. Its icon, `F3`, and `Ctrl`/`Cmd`+`K` all open the same permission-aware search overlay; the titlebar does not reserve a persistent query field. | Accepted | Core shell and Search WebUI |
+| UX-034 | Every headed `PageLayout` declares one of `overview`, `collection`, `detail`, `editor`, or `workspace` independently from its standalone/workspace/embedded geometry. Its actions use the matching semantic `PageActionBar`: refreshable pages provide Reload in the right-aligned trailing group immediately before Create/primary actions; collections keep Create far right; read-only pages do not invent Save. The trailing placement supersedes the earlier leading-Reload rule (2026-09-07, Core #295). | Accepted | Core and all module WebUIs |
+| UX-035 | Editor action bars expose clean, dirty, and saving state; always retain Discard immediately before the far-right Save; centrally disable both while clean or saving; and participate in the unsaved-change navigation guard. Danger actions occupy the explicit separated destructive group after ordinary actions and before editor persistence. | Accepted | Core and all module WebUIs |
 
 ## Confirmed Implementation Decisions
 
@@ -161,14 +165,19 @@ Decision: the WebUI shell exposes a small, stable appearance contract based on
 shared CSS tokens and persisted user preference selection.
 
 - Core applies `system`, `light`, and `dark` preferences at the document root.
+- Core applies validated user accent presets through `data-palette`; palette
+  values change semantic tokens globally and never require module CSS changes.
 - Core owns shared tokens such as `--bg`, `--bar`, `--panel`, `--surface`,
   `--line`, `--line-dark`, `--text`, `--text-strong`, `--muted`, semantic
   status colors, radii, shadows, and disabled-control colors.
 - Modules must style new UI with these tokens and shared controls. Module-local
   CSS may tune layout and spacing, but it must not introduce a separate
   appearance system.
-- Appearance controls live in user settings first. Tenant defaults and policy
-  enforcement can be added later without changing the token contract.
+- Appearance controls live in user settings. A personal palette wins over
+  unlocked tenant and system defaults; system and tenant locks take precedence.
+  Advanced personal token overrides additionally require system opt-in and may
+  be narrowed by tenant policy. Their versioned import/export document is
+  validated and applied all-or-nothing in both light and dark modes.
 - Visual preview in settings is illustrative; it must reflect token families,
   not become a second theme implementation.
 
@@ -227,7 +236,9 @@ instead of reproducing their behavior.
   not self-explanatory.
 - `help` content is contextual guidance, not the accessible name. The persisted
   `show_inline_help_hints` user preference hides only the `InlineHelp` marker by
-  applying `ui-hide-help-hints` at the document root.
+  applying `ui-hide-help-hints` at the document root. When shown, the shared
+  marker is a labelled, keyboard-focusable help control and exposes its tooltip
+  on focus as well as pointer hover.
 - Shared action-bearing components accept an optional disabled reason. In
   particular, `MailServerSettingsPanel` forwards protocol-specific test
   blockers into the shared focusable disabled-action tooltip; modules provide
@@ -319,7 +330,7 @@ converted or reviewed.
 | Configuration packages | `govoplan-admin` | Catalog/import work exists, but package editing can still drift toward technical fields. | Add guided import/review/problem-list flow. |
 | Retention and privacy | `govoplan-core` | Typed effective-policy editor exposes source paths, narrowing semantics, platform locks, permission/target blockers, and explicit clean/loading/save states. | Broader governed-change review remains module-owned where a policy change requires approval. |
 | API keys | `govoplan-access` / admin UI | Security-sensitive creation needs least-privilege guidance. | Add scoped creation wizard with expiry/owner review. |
-| User settings | `govoplan-core` | Simple typed sections use unsaved-change guards, quiet result feedback, contextual help, and explicit busy/clean disabled-action reasons. | Keep bounded; new contributed sections must satisfy the checklist. |
+| User settings | `govoplan-core` | Simple typed sections use unsaved-change guards, quiet result feedback, contextual help, explicit busy/clean disabled-action reasons, and an effective appearance source. Palette selection and light/dark preview are shared with system and tenant administration. | Keep bounded; new contributed sections must satisfy the checklist. |
 
 ## Impact Index
 
@@ -335,7 +346,7 @@ converted or reviewed.
 | Automation/workflow commands | Hidden side effects would undermine accountability. | Action/effect preview, system-actor display, command record, retry/quarantine/manual states, and audit links. |
 | Postbox and encrypted communication | Retraction and access can be misunderstood. | Honest key-fetch/decryption state, expiry limits, recipient/device access provenance, and delivery evidence. |
 | API keys | Security-sensitive creation and scope selection. | Scoped creation wizard, least-privilege suggestions, clear expiry/owner explanation. |
-| User settings | Needs clarity and persistence across profile/interface/preferences. | Simple settings sections with immediate feedback and no double-click navigation traps. |
+| User settings | Needs clarity and persistence across profile/interface/preferences. | Simple settings sections with immediate feedback, explicit inherit/reset semantics, effective-source provenance, and no double-click navigation traps. |
 
 ## Review Checklist
 

docs/WEBUI_BUNDLE_BUDGETS.md

@@ -59,6 +59,14 @@ The first budgeted full-product build reported:
 
 ## Verification
 
+Development startup explicitly prebundles the Excel reader's browser/universal
+entrypoints and the lazy rich-text editor's Tiptap dependencies. These are
+Core-installed vendor dependencies, not eager optional-module imports. This
+avoids first-time Campaign/Template navigation triggering a second dependency
+optimization and page reload. Module descriptors and pages remain lazy, and
+production bundle budgets remain unchanged. The Core interface-pattern check
+verifies this include list and keeps optional GovOPlaN modules excluded.
+
 ```bash
 cd /mnt/DATA/git/govoplan-core/webui
 npm run build
@@ -69,3 +77,33 @@ npm run test:module-permutations
 The build gate also catches accidental eager imports: a page pulled into the
 entry closure consumes the initial budget, while an oversized page or module
 descriptor consumes the asynchronous chunk budget.
+
+The startup shell imports appearance validation/application from the pure
+`appearanceOverrides.ts` runtime. Settings-only color controls, JSON import/export,
+and previews remain in `AppearanceOverridesEditor.tsx` behind the existing lazy
+Settings route. Importing a runtime helper from a module that also owns editor
+components can accidentally pull the entire editor into the startup chunk.
+Public helper exports remain compatible; theme application is still synchronous.
+The versioned default color document and its deep-clone helper live in
+`appearanceOverrideDefaults.ts`, loaded with that editor. Applying saved overrides
+does not load editor defaults or construct a draft. The default values and public
+helper names are unchanged; the theme regression checks independent draft clones
+as well as synchronous validation, application, and reset.
+
+`PasswordField` keeps ordinary input and reveal controls synchronous. Its
+optional `PasswordGeneratorDialog` is imported only after an enabled, editable
+generator is explicitly opened, not for every sign-in/password field. Loading
+and failures use the shared resource boundary; the underlying field remains
+usable. Closing or revoking generation while loading cannot apply a candidate.
+The secure browser RNG, generation policy, public exports, and explicit
+"Use password" confirmation remain unchanged. The isolated browser fixture
+does not import the Core barrel, so it can verify that the generator is not
+requested before opening it, along with cancel/use and focus restoration.
+
+Deutsch: Normale Passworteingabe und Sichtbarkeitssteuerung bleiben unmittelbar
+verfügbar. Der optionale Generator wird erst beim bewussten Öffnen eines
+aktivierten, bearbeitbaren Felds geladen; Lade- und Fehlerzustände nutzen die
+gemeinsame Ressourcenanzeige. Ohne "Passwort verwenden" wird kein Kandidat
+übernommen. Sichere Browser-Zufallszahlen, Richtlinien und öffentliche
+Schnittstellen bleiben unverändert. Wird die Generierung während des Ladens
+deaktiviert, öffnet eine verspätete Antwort keinen Dialog.

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "govoplan-core"
-version = "0.1.15"
+version = "0.1.46"
 description = "Reusable GovOPlaN platform core, access, tenancy, and RBAC components."
 readme = "README.md"
 requires-python = ">=3.12"
@@ -15,7 +15,7 @@ dependencies = [
   "fastapi>=0.139,<1",
   "pydantic>=2,<3",
   "pydantic-settings>=2,<3",
-  "cryptography>=48.0.1,<50",
+  "cryptography>=50.0.0,<51",
   "celery>=5,<6",
   "redis>=5,<6",
   "alembic>=1,<2",
@@ -26,7 +26,7 @@ dependencies = [
 where = ["src"]
 
 [tool.setuptools.package-data]
-govoplan_core = ["py.typed"]
+govoplan_core = ["py.typed", "resources/*.json"]
 
 [tool.setuptools.data-files]
 "govoplan_core_runtime" = ["alembic.ini"]

src/govoplan_core/admin/models.py

@@ -12,7 +12,7 @@ class SystemSettings(Base, TimestampMixin):
     __tablename__ = "core_system_settings"
 
     id: Mapped[str] = mapped_column(String(36), primary_key=True, default="global")
-    default_locale: Mapped[str] = mapped_column(String(20), default="en", nullable=False)
+    default_locale: Mapped[str] = mapped_column(String(20), default="de", nullable=False)
     allow_tenant_custom_groups: Mapped[bool] = mapped_column(Boolean, default=True, nullable=False)
     allow_tenant_custom_roles: Mapped[bool] = mapped_column(Boolean, default=True, nullable=False)
     allow_tenant_api_keys: Mapped[bool] = mapped_column(Boolean, default=True, nullable=False)
@@ -20,4 +20,3 @@ class SystemSettings(Base, TimestampMixin):
 
 
 __all__ = ["SystemSettings"]
-

src/govoplan_core/api/v1/schemas.py

@@ -3,7 +3,9 @@ from __future__ import annotations
 from datetime import datetime
 from typing import Any, Literal
 
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+from govoplan_core.core.appearance import normalize_appearance_overrides
 
 
 class AuditLogItemResponse(BaseModel):
@@ -84,7 +86,7 @@ class TenantInfo(BaseModel):
     slug: str
     name: str
     is_active: bool = True
-    default_locale: str = "en"
+    default_locale: str = "de"
     enabled_language_codes: list[str] = Field(default_factory=list)
 
 
@@ -93,6 +95,53 @@ class TenantMembershipInfo(TenantInfo):
     is_active: bool = True
 
 
+class NavigationSeparatorPayload(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    id: str = Field(pattern=r"^separator:[a-zA-Z0-9_.:-]+$", max_length=255)
+    label: str = Field(default="", max_length=120, pattern=r"^[^\x00-\x1f]*$")
+
+
+class NavigationPreferencesPayload(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    contract_version: Literal["1"] = "1"
+    order: list[str] = Field(default_factory=list, max_length=256)
+    hidden: list[str] = Field(default_factory=list, max_length=256)
+    locked: list[str] = Field(default_factory=list, max_length=256)
+    separators: list[NavigationSeparatorPayload] | None = Field(default=None, max_length=256)
+
+
+class AppearanceModeOverrides(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    accent: str
+    accent_foreground: str
+    surface: str
+    surface_foreground: str
+    success: str
+    success_foreground: str
+    info: str
+    info_foreground: str
+    warning: str
+    warning_foreground: str
+    danger: str
+    danger_foreground: str
+
+
+class AppearanceOverridesDocument(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    schema_version: Literal["1"] = "1"
+    light: AppearanceModeOverrides
+    dark: AppearanceModeOverrides
+
+    @model_validator(mode="after")
+    def validate_accessibility(self) -> "AppearanceOverridesDocument":
+        normalize_appearance_overrides(self.model_dump(mode="json"))
+        return self
+
+
 class UserUiPreferences(BaseModel):
     model_config = ConfigDict(extra="ignore")
 
@@ -101,6 +150,20 @@ class UserUiPreferences(BaseModel):
     reduce_motion: bool = False
     sticky_section_sidebars: bool = True
     theme: Literal["system", "light", "dark"] = "system"
+    palette: Literal["default", "civic_blue", "forest", "plum"] | None = None
+    appearance_overrides: AppearanceOverridesDocument | None = None
+    navigation: NavigationPreferencesPayload | None = None
+
+
+class EffectiveAppearanceInfo(BaseModel):
+    palette: Literal["default", "civic_blue", "forest", "plum"] = "default"
+    source: Literal["user", "tenant", "system", "tenant_lock", "system_lock"] = "system"
+    locked: bool = False
+    system_default_palette: Literal["default", "civic_blue", "forest", "plum"] = "default"
+    tenant_default_palette: Literal["default", "civic_blue", "forest", "plum"] | None = None
+    inherited_palette: Literal["default", "civic_blue", "forest", "plum"] = "default"
+    custom_overrides: AppearanceOverridesDocument | None = None
+    custom_overrides_allowed: bool = False
 
 
 class UserInfo(BaseModel):
@@ -113,9 +176,12 @@ class UserInfo(BaseModel):
     tenant_display_name: str | None = None
     is_tenant_admin: bool = False
     password_reset_required: bool = False
+    required_auth_action: Literal["change_password"] | None = None
+    local_password: bool = False
     preferred_language: str | None = None
     enabled_language_codes: list[str] = Field(default_factory=list)
     ui_preferences: UserUiPreferences = Field(default_factory=UserUiPreferences)
+    appearance: EffectiveAppearanceInfo = Field(default_factory=EffectiveAppearanceInfo)
 
 
 class AuthSessionUserInfo(BaseModel):
@@ -126,6 +192,8 @@ class AuthSessionUserInfo(BaseModel):
     tenant_display_name: str | None = None
     is_tenant_admin: bool = False
     password_reset_required: bool = False
+    required_auth_action: Literal["change_password"] | None = None
+    local_password: bool = False
 
 
 class AuthSessionResponse(BaseModel):
@@ -210,7 +278,7 @@ class AuthProfileResponse(BaseModel):
     active_tenant: TenantInfo
     available_languages: list[LanguageInfo] = Field(default_factory=list)
     enabled_language_codes: list[str] = Field(default_factory=list)
-    default_language: str = "en"
+    default_language: str = "de"
     profile_loaded: bool = True
 
 
@@ -239,7 +307,7 @@ class LoginResponse(BaseModel):
     principal: PrincipalContextInfo | None = None
     available_languages: list[LanguageInfo] = Field(default_factory=list)
     enabled_language_codes: list[str] = Field(default_factory=list)
-    default_language: str = "en"
+    default_language: str = "de"
     profile_loaded: bool = True
     roles_loaded: bool = True
     groups_loaded: bool = True
@@ -257,7 +325,7 @@ class MeResponse(BaseModel):
     principal: PrincipalContextInfo | None = None
     available_languages: list[LanguageInfo] = Field(default_factory=list)
     enabled_language_codes: list[str] = Field(default_factory=list)
-    default_language: str = "en"
+    default_language: str = "de"
     profile_loaded: bool = True
     roles_loaded: bool = True
     groups_loaded: bool = True

src/govoplan_core/celery_app.py

@@ -18,7 +18,9 @@ from celery.signals import (
 
 from govoplan_core.core.campaigns import (
     CAPABILITY_CAMPAIGNS_DELIVERY_TASKS,
+    CAPABILITY_CAMPAIGNS_SCHEDULES,
     CampaignDeliveryTaskProvider,
+    CampaignScheduleProvider,
 )
 from govoplan_core.core.calendar import (
     CAPABILITY_CALENDAR_OUTBOX,
@@ -100,6 +102,7 @@ celery.conf.update(
     task_routes={
         "govoplan.campaigns.send_email": {"queue": "send_email"},
         "govoplan.campaigns.append_sent": {"queue": "append_sent"},
+        "govoplan.campaigns.dispatch_schedules": {"queue": "default"},
         "govoplan.notifications.deliver": {"queue": "notifications"},
         "govoplan.notifications.deliver_pending": {"queue": "notifications"},
         "govoplan.mail.dispatch_outbox": {"queue": "mail"},
@@ -132,6 +135,11 @@ celery.conf.update(
             "schedule": 60.0,
             "args": (None, 100),
         },
+        "campaign-schedules-every-minute": {
+            "task": "govoplan.campaigns.dispatch_schedules",
+            "schedule": 60.0,
+            "args": (None, 50),
+        },
         "mail-outbox-every-five-seconds": {
             "task": "govoplan.mail.dispatch_outbox",
             "schedule": 5.0,
@@ -555,6 +563,18 @@ def _campaign_delivery_tasks(
     return capability
 
 
+def _campaign_schedules(
+    registry: PlatformRegistry | None = None,
+) -> CampaignScheduleProvider | None:
+    registry = registry or _platform_registry()
+    if not registry.has_capability(CAPABILITY_CAMPAIGNS_SCHEDULES):
+        return None
+    capability = registry.require_capability(CAPABILITY_CAMPAIGNS_SCHEDULES)
+    if not isinstance(capability, CampaignScheduleProvider):
+        raise RuntimeError("Campaign schedule capability is invalid")
+    return capability
+
+
 def _notification_dispatch(
     registry: PlatformRegistry | None = None,
 ) -> NotificationDispatchProvider:
@@ -699,13 +719,70 @@ def _idm_assignment_lifecycle(
     return capability
 
 
+@celery.task(
+    name="govoplan.campaigns.dispatch_schedules",
+    bind=True,
+    max_retries=0,
+)
+def dispatch_campaign_schedules(
+    self,
+    tenant_id: str | None = None,
+    limit: int = 50,
+):
+    """Prepare manual drafts or governed autonomous Mail commands for due schedules."""
+
+    from govoplan_core.db.session import get_database
+
+    with get_database().SessionLocal() as session:
+        registry = _platform_registry()
+        defaults = {
+            "selected": 0,
+            "prepared": 0,
+            "autonomous_prepared": 0,
+            "failed": 0,
+            "completed": 0,
+            "coalesced": 0,
+            "duplicates": 0,
+            "deferred": 0,
+            "campaign_ids": [],
+            "operator_actions": [],
+            "refreshed": {
+                "checked": 0,
+                "accepted": 0,
+                "uncertain": 0,
+                "failed": 0,
+                "skipped": 0,
+            },
+        }
+        if not registry.has_capability(CAPABILITY_CAMPAIGNS_SCHEDULES):
+            return defaults
+        result = _run_tenant_worker_batches(
+            registry,
+            session,
+            capability_name=CAPABILITY_CAMPAIGNS_SCHEDULES,
+            tenant_id=tenant_id,
+            operation=lambda effective_tenant_id: _campaign_schedules(
+                registry
+            ).dispatch_due(  # type: ignore[union-attr]
+                session,
+                tenant_id=effective_tenant_id,
+                limit=limit,
+            ),
+            defaults=defaults,
+            work_state="new",
+        )
+        session.commit()
+        return result
+
+
 @celery.task(name="govoplan.campaigns.send_email", bind=True, max_retries=0)
 def send_email(self, job_id: str):
     """Send one explicitly queued campaign job.
 
     SMTP failures are persisted but are not retried implicitly. A worker-loss
-    redelivery is safe because the delivery service converts an unfinished
-    SMTP attempt into ``outcome_unknown`` instead of transmitting again.
+    redelivery leaves an active delivery claim unchanged instead of transmitting
+    again. Explicit fenced recovery requires stopped-owner evidence before an
+    abandoned attempt can become ``outcome_unknown`` for reconciliation.
     """
 
     from govoplan_core.db.session import get_database
@@ -1122,7 +1199,7 @@ def dispatch_postbox_routes(
     tenant_id: str | None = None,
     limit: int = 50,
 ):
-    """Deliver due Postbox vacancy escalations from durable route rows."""
+    """Deliver due routes and reconcile assignment-derived notification facts."""
 
     from govoplan_core.db.session import get_database
 
@@ -1136,21 +1213,40 @@ def dispatch_postbox_routes(
             "cancelled": 0,
             "failed": 0,
             "route_ids": [],
+            "lifecycle_scanned": 0,
+            "lifecycle_changed": 0,
+            "lifecycle_events": 0,
+            "lifecycle_notifications": 0,
+            "lifecycle_notification_failures": 0,
         }
         if not registry.has_capability(CAPABILITY_POSTBOX_ROUTING):
             return defaults
+        provider = _postbox_routing_provider(registry)
+
+        def dispatch_tenant(effective_tenant_id: str) -> Mapping[str, object]:
+            route_result = provider.dispatch_due_routes(  # type: ignore[union-attr]
+                session,
+                tenant_id=effective_tenant_id,
+                limit=limit,
+            )
+            lifecycle_result = provider.reconcile_notification_lifecycle(  # type: ignore[union-attr]
+                session,
+                tenant_id=effective_tenant_id,
+                limit=limit,
+            )
+            return {
+                **route_result,
+                **{
+                    f"lifecycle_{key}": value for key, value in lifecycle_result.items()
+                },
+            }
+
         result = _run_tenant_worker_batches(
             registry,
             session,
             capability_name=CAPABILITY_POSTBOX_ROUTING,
             tenant_id=tenant_id,
-            operation=lambda effective_tenant_id: _postbox_routing_provider(
-                registry
-            ).dispatch_due_routes(  # type: ignore[union-attr]
-                session,
-                tenant_id=effective_tenant_id,
-                limit=limit,
-            ),
+            operation=dispatch_tenant,
             defaults=defaults,
         )
         session.commit()

src/govoplan_core/core/access.py

@@ -24,6 +24,10 @@ CAPABILITY_ACCESS_TENANT_PROVISIONER = "access.tenantProvisioner"
 CAPABILITY_ACCESS_FIRST_ADMIN_PROVISIONER = "access.firstAdminProvisioner"
 CAPABILITY_ACCESS_ADMINISTRATION = "access.administration"
 CAPABILITY_ACCESS_GOVERNANCE_MATERIALIZER = "access.governanceMaterializer"
+CAPABILITY_ACCESS_GOVERNANCE_PROJECTION_V1 = "access.governanceProjection.v1"
+CAPABILITY_POLICY_ACCESS_EXPLANATION_SUBJECTS = (
+    "policy.access_explanation_subjects"
+)
 CAPABILITY_TENANCY_TENANT_RESOLVER = "tenancy.tenantResolver"
 CAPABILITY_AUDIT_SINK = "audit.sink"
 CAPABILITY_AUDIT_RECORDER = "audit.recorder"
@@ -49,6 +53,7 @@ ACCESS_CAPABILITY_NAMES = frozenset(
         CAPABILITY_ACCESS_FIRST_ADMIN_PROVISIONER,
         CAPABILITY_ACCESS_ADMINISTRATION,
         CAPABILITY_ACCESS_GOVERNANCE_MATERIALIZER,
+        CAPABILITY_ACCESS_GOVERNANCE_PROJECTION_V1,
         CAPABILITY_TENANCY_TENANT_RESOLVER,
         CAPABILITY_AUDIT_SINK,
         CAPABILITY_AUDIT_RECORDER,
@@ -177,6 +182,15 @@ class PrincipalRef:
         )
 
 
+@dataclass(frozen=True, slots=True)
+class AccessExplanationSubjectDecision:
+    allow_other_users: bool
+    reason: str
+    source: str
+    required_scope: str | None = None
+    provenance: Mapping[str, object] = field(default_factory=dict)
+
+
 def _optional_str(value: object | None) -> str | None:
     return str(value) if value is not None else None
 
@@ -378,6 +392,82 @@ class GovernanceTemplateMaterialization:
     required: bool = False
 
 
+GovernanceProjectionOperation = Literal["upsert", "remove"]
+GovernanceProjectionStatus = Literal[
+    "created",
+    "updated",
+    "unchanged",
+    "removed",
+    "absent",
+    "blocked",
+    "failed",
+]
+
+
+@dataclass(frozen=True, slots=True)
+class GovernanceProjectionCommand:
+    """Stable Access-owned input for one governance assignment projection."""
+
+    assignment_id: str
+    operation: GovernanceProjectionOperation
+    template: GovernanceTemplateMaterialization
+    provenance: Mapping[str, str] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        if not self.assignment_id or len(self.assignment_id) > 255:
+            raise ValueError("Governance projection assignment ids must contain at most 255 characters.")
+        if len(self.provenance) > 20:
+            raise ValueError("Governance projection provenance supports at most 20 entries.")
+        for key, value in self.provenance.items():
+            if not key or len(key) > 100 or len(value) > 500:
+                raise ValueError("Governance projection provenance entries exceed their bounds.")
+
+
+@dataclass(frozen=True, slots=True)
+class GovernanceProjectionBatch:
+    """Versioned, bounded reconciliation request independent of Admin internals."""
+
+    operation_id: str
+    commands: tuple[GovernanceProjectionCommand, ...]
+    version: Literal["1"] = "1"
+    dry_run: bool = False
+
+    def __post_init__(self) -> None:
+        if not self.operation_id or len(self.operation_id) > 255:
+            raise ValueError("Governance projection operation ids must contain at most 255 characters.")
+        if not self.commands or len(self.commands) > 500:
+            raise ValueError("Governance projection batches must contain between 1 and 500 commands.")
+        assignment_ids = [command.assignment_id for command in self.commands]
+        if len(assignment_ids) != len(set(assignment_ids)):
+            raise ValueError("Governance projection assignment ids must be unique within a batch.")
+
+
+@dataclass(frozen=True, slots=True)
+class GovernanceProjectionOutcome:
+    assignment_id: str
+    template_id: str
+    tenant_id: str
+    kind: Literal["group", "role"]
+    operation: GovernanceProjectionOperation
+    status: GovernanceProjectionStatus
+    resource_id: str | None = None
+    blocker_codes: tuple[str, ...] = ()
+    message: str | None = None
+    provenance: Mapping[str, str] = field(default_factory=dict)
+
+
+@dataclass(frozen=True, slots=True)
+class GovernanceProjectionResult:
+    operation_id: str
+    outcomes: tuple[GovernanceProjectionOutcome, ...]
+    version: Literal["1"] = "1"
+    dry_run: bool = False
+
+    @property
+    def blocked(self) -> tuple[GovernanceProjectionOutcome, ...]:
+        return tuple(item for item in self.outcomes if item.status in {"blocked", "failed"})
+
+
 @dataclass(frozen=True, slots=True)
 class AuditEvent:
     event_type: str
@@ -564,6 +654,18 @@ class AccessExplanationService(Protocol):
         ...
 
 
+@runtime_checkable
+class AccessExplanationSubjectPolicy(Protocol):
+    def decide_subject_selection(
+        self,
+        session: object,
+        principal: PrincipalRef,
+        *,
+        tenant_id: str,
+    ) -> AccessExplanationSubjectDecision:
+        ...
+
+
 @runtime_checkable
 class TenantAccessProvisioner(Protocol):
     def ensure_default_roles(self, session: object, tenant: object | None = None) -> Mapping[str, object]:
@@ -669,6 +771,18 @@ class AccessGovernanceMaterializer(Protocol):
         ...
 
 
+@runtime_checkable
+class AccessGovernanceProjectionV1(Protocol):
+    """Bulk reconciliation boundary for Admin-owned governance assignments."""
+
+    def reconcile(
+        self,
+        session: object,
+        batch: GovernanceProjectionBatch,
+    ) -> GovernanceProjectionResult:
+        ...
+
+
 @runtime_checkable
 class AuditSink(Protocol):
     def record(self, event: AuditEvent) -> None:

src/govoplan_core/core/appearance.py

@@ -0,0 +1,223 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+import re
+from typing import Any, Literal, Mapping
+
+
+AppearancePalette = Literal["default", "civic_blue", "forest", "plum"]
+AppearanceSource = Literal["user", "tenant", "system", "tenant_lock", "system_lock"]
+
+APPEARANCE_PALETTES: tuple[AppearancePalette, ...] = ("default", "civic_blue", "forest", "plum")
+APPEARANCE_SETTINGS_KEY = "appearance"
+APPEARANCE_OVERRIDE_SCHEMA_VERSION = "1"
+APPEARANCE_OVERRIDE_TOKENS: tuple[str, ...] = (
+    "accent", "accent_foreground", "surface", "surface_foreground",
+    "success", "success_foreground", "info", "info_foreground",
+    "warning", "warning_foreground", "danger", "danger_foreground",
+)
+_STATUS_TOKENS = ("success", "info", "warning", "danger")
+_HEX_COLOR = re.compile(r"^#[0-9a-fA-F]{6}$")
+
+
+@dataclass(frozen=True, slots=True)
+class EffectiveAppearance:
+    palette: AppearancePalette
+    source: AppearanceSource
+    locked: bool
+    system_default_palette: AppearancePalette
+    tenant_default_palette: AppearancePalette | None
+    inherited_palette: AppearancePalette
+    custom_overrides: dict[str, object] | None = None
+    custom_overrides_allowed: bool = False
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "palette": self.palette,
+            "source": self.source,
+            "locked": self.locked,
+            "system_default_palette": self.system_default_palette,
+            "tenant_default_palette": self.tenant_default_palette,
+            "inherited_palette": self.inherited_palette,
+            "custom_overrides": self.custom_overrides,
+            "custom_overrides_allowed": self.custom_overrides_allowed,
+        }
+
+
+def normalize_appearance_palette(value: object, *, fallback: AppearancePalette | None = None) -> AppearancePalette | None:
+    normalized = str(value or "").strip().lower()
+    return normalized if normalized in APPEARANCE_PALETTES else fallback  # type: ignore[return-value]
+
+
+def appearance_settings(settings: Mapping[str, Any] | None) -> tuple[AppearancePalette | None, bool]:
+    raw = settings.get(APPEARANCE_SETTINGS_KEY) if isinstance(settings, Mapping) else None
+    if not isinstance(raw, Mapping):
+        return None, False
+    return normalize_appearance_palette(raw.get("default_palette")), raw.get("palette_locked") is True
+
+
+def appearance_custom_overrides_policy(settings: Mapping[str, Any] | None) -> bool | None:
+    raw = settings.get(APPEARANCE_SETTINGS_KEY) if isinstance(settings, Mapping) else None
+    if not isinstance(raw, Mapping) or "allow_custom_overrides" not in raw:
+        return None
+    return raw.get("allow_custom_overrides") is True
+
+
+def update_appearance_custom_overrides_policy(
+    settings: Mapping[str, Any] | None,
+    *,
+    allowed: bool | None,
+) -> dict[str, Any]:
+    updated = dict(settings or {})
+    appearance = dict(updated.get(APPEARANCE_SETTINGS_KEY) or {}) if isinstance(updated.get(APPEARANCE_SETTINGS_KEY), Mapping) else {}
+    if allowed is None:
+        appearance.pop("allow_custom_overrides", None)
+    else:
+        appearance["allow_custom_overrides"] = allowed
+    if appearance:
+        updated[APPEARANCE_SETTINGS_KEY] = appearance
+    else:
+        updated.pop(APPEARANCE_SETTINGS_KEY, None)
+    return updated
+
+
+def normalize_appearance_overrides(value: object) -> dict[str, object] | None:
+    """Validate and canonicalize the versioned, all-or-nothing color contract."""
+
+    if value is None:
+        return None
+    if not isinstance(value, Mapping):
+        raise ValueError("Appearance overrides must be an object.")
+    if set(value) != {"schema_version", "light", "dark"}:
+        raise ValueError("Appearance overrides must contain only schema_version, light, and dark.")
+    if str(value.get("schema_version")) != APPEARANCE_OVERRIDE_SCHEMA_VERSION:
+        raise ValueError("Unsupported appearance override schema version.")
+    normalized: dict[str, object] = {"schema_version": APPEARANCE_OVERRIDE_SCHEMA_VERSION}
+    for mode in ("light", "dark"):
+        raw_mode = value.get(mode)
+        if not isinstance(raw_mode, Mapping) or set(raw_mode) != set(APPEARANCE_OVERRIDE_TOKENS):
+            raise ValueError(f"Appearance override mode {mode} must define every supported token exactly once.")
+        colors: dict[str, str] = {}
+        for token in APPEARANCE_OVERRIDE_TOKENS:
+            color = str(raw_mode.get(token) or "").strip().lower()
+            if not _HEX_COLOR.fullmatch(color):
+                raise ValueError(f"Appearance override {mode}.{token} must be a six-digit hexadecimal color.")
+            colors[token] = color
+        _validate_mode_accessibility(mode, colors)
+        normalized[mode] = colors
+    return normalized
+
+
+def _validate_mode_accessibility(mode: str, colors: Mapping[str, str]) -> None:
+    pairs = (
+        ("accent", "accent_foreground"), ("surface", "surface_foreground"),
+        ("success", "success_foreground"), ("info", "info_foreground"),
+        ("warning", "warning_foreground"), ("danger", "danger_foreground"),
+    )
+    for background, foreground in pairs:
+        if _contrast_ratio(colors[background], colors[foreground]) < 4.5:
+            raise ValueError(f"Appearance override {mode}.{foreground} must have WCAG AA contrast against {mode}.{background}.")
+    status_colors = [colors[token] for token in _STATUS_TOKENS]
+    for index, first in enumerate(status_colors):
+        for second in status_colors[index + 1:]:
+            if _rgb_distance(first, second) < 12:
+                raise ValueError(f"Appearance override status colors in {mode} must remain visibly distinct.")
+
+
+def _relative_luminance(color: str) -> float:
+    channels = [int(color[index:index + 2], 16) / 255 for index in (1, 3, 5)]
+    linear = [channel / 12.92 if channel <= 0.04045 else ((channel + 0.055) / 1.055) ** 2.4 for channel in channels]
+    return 0.2126 * linear[0] + 0.7152 * linear[1] + 0.0722 * linear[2]
+
+
+def _contrast_ratio(first: str, second: str) -> float:
+    high, low = sorted((_relative_luminance(first), _relative_luminance(second)), reverse=True)
+    return (high + 0.05) / (low + 0.05)
+
+
+def _rgb_distance(first: str, second: str) -> float:
+    first_channels = [int(first[index:index + 2], 16) for index in (1, 3, 5)]
+    second_channels = [int(second[index:index + 2], 16) for index in (1, 3, 5)]
+    return sum((left - right) ** 2 for left, right in zip(first_channels, second_channels, strict=True)) ** 0.5
+
+
+def update_appearance_settings(
+    settings: Mapping[str, Any] | None,
+    *,
+    default_palette: AppearancePalette | None,
+    palette_locked: bool,
+) -> dict[str, Any]:
+    updated = dict(settings or {})
+    appearance = dict(updated.get(APPEARANCE_SETTINGS_KEY) or {}) if isinstance(updated.get(APPEARANCE_SETTINGS_KEY), Mapping) else {}
+    if default_palette is None:
+        appearance.pop("default_palette", None)
+    else:
+        normalized = normalize_appearance_palette(default_palette)
+        if normalized is None:
+            raise ValueError("Unsupported appearance palette.")
+        appearance["default_palette"] = normalized
+    if palette_locked:
+        appearance["palette_locked"] = True
+    else:
+        appearance.pop("palette_locked", None)
+    if appearance:
+        updated[APPEARANCE_SETTINGS_KEY] = appearance
+    else:
+        updated.pop(APPEARANCE_SETTINGS_KEY, None)
+    return updated
+
+
+def resolve_effective_appearance(
+    *,
+    system_settings: Mapping[str, Any] | None,
+    tenant_settings: Mapping[str, Any] | None,
+    user_settings: Mapping[str, Any] | None,
+) -> EffectiveAppearance:
+    system_palette, system_locked = appearance_settings(system_settings)
+    system_palette = system_palette or "default"
+    tenant_palette, tenant_locked = appearance_settings(tenant_settings)
+    inherited_palette = tenant_palette or system_palette
+    raw_ui = user_settings.get("ui") if isinstance(user_settings, Mapping) else None
+    user_palette = normalize_appearance_palette(raw_ui.get("palette")) if isinstance(raw_ui, Mapping) else None
+    system_custom_policy = appearance_custom_overrides_policy(system_settings) is True
+    tenant_custom_policy = appearance_custom_overrides_policy(tenant_settings)
+    custom_overrides_allowed = system_custom_policy and tenant_custom_policy is not False and not system_locked and not tenant_locked
+    try:
+        custom_overrides = normalize_appearance_overrides(raw_ui.get("appearance_overrides")) if isinstance(raw_ui, Mapping) else None
+    except ValueError:
+        custom_overrides = None
+    if not custom_overrides_allowed:
+        custom_overrides = None
+
+    if system_locked:
+        return EffectiveAppearance(system_palette, "system_lock", True, system_palette, tenant_palette, system_palette)
+    if tenant_locked:
+        return EffectiveAppearance(inherited_palette, "tenant_lock", True, system_palette, tenant_palette, inherited_palette)
+    return EffectiveAppearance(
+        user_palette or inherited_palette,
+        "user" if user_palette else "tenant" if tenant_palette else "system",
+        False,
+        system_palette,
+        tenant_palette,
+        inherited_palette,
+        custom_overrides,
+        custom_overrides_allowed,
+    )
+
+
+__all__ = [
+    "APPEARANCE_PALETTES",
+    "APPEARANCE_SETTINGS_KEY",
+    "APPEARANCE_OVERRIDE_SCHEMA_VERSION",
+    "APPEARANCE_OVERRIDE_TOKENS",
+    "AppearancePalette",
+    "AppearanceSource",
+    "EffectiveAppearance",
+    "appearance_settings",
+    "appearance_custom_overrides_policy",
+    "normalize_appearance_overrides",
+    "normalize_appearance_palette",
+    "resolve_effective_appearance",
+    "update_appearance_settings",
+    "update_appearance_custom_overrides_policy",
+]

src/govoplan_core/core/application_status.py

@@ -0,0 +1,81 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from datetime import datetime
+from typing import Protocol, runtime_checkable
+
+
+CAPABILITY_APPLICATION_STATUS_PROJECTION = "application_status.projection"
+
+
+@runtime_checkable
+class ApplicationStatusProjectionProvider(Protocol):
+    """Bounded applicant-status access without exposing the owning module's data."""
+
+    def tenant_id_for_tracking_id(
+        self,
+        session: object,
+        *,
+        tracking_id: str,
+    ) -> str | None:
+        ...
+
+    def public_access_challenge(
+        self,
+        session: object,
+        *,
+        tracking_id: str,
+    ) -> Mapping[str, object]:
+        ...
+
+    def get_authenticated_projection(
+        self,
+        session: object,
+        principal: object,
+        *,
+        tracking_id: str,
+        observed_at: datetime,
+    ) -> Mapping[str, object]:
+        ...
+
+    def get_public_projection(
+        self,
+        session: object,
+        *,
+        tracking_id: str,
+        token: str | None,
+        observed_at: datetime,
+    ) -> Mapping[str, object]:
+        ...
+
+    def request_email_link(
+        self,
+        session: object,
+        *,
+        tracking_id: str,
+        email: str,
+        requested_at: datetime,
+    ) -> bool:
+        ...
+
+
+def application_status_projection_provider(
+    registry: object | None,
+) -> ApplicationStatusProjectionProvider | None:
+    if registry is None or not hasattr(registry, "has_capability"):
+        return None
+    if not registry.has_capability(CAPABILITY_APPLICATION_STATUS_PROJECTION):
+        return None
+    capability = registry.capability(CAPABILITY_APPLICATION_STATUS_PROJECTION)
+    return (
+        capability
+        if isinstance(capability, ApplicationStatusProjectionProvider)
+        else None
+    )
+
+
+__all__ = [
+    "ApplicationStatusProjectionProvider",
+    "CAPABILITY_APPLICATION_STATUS_PROJECTION",
+    "application_status_projection_provider",
+]

src/govoplan_core/core/calendar.py

@@ -45,6 +45,15 @@ class CalendarEventRef:
     outbox_operation_id: str | None = None
 
 
+@dataclass(frozen=True, slots=True)
+class CalendarEventReleaseRef:
+    event_id: str
+    accepted: bool = True
+    already_released: bool = False
+    external_state: str = "local_released"
+    outbox_operation_id: str | None = None
+
+
 @dataclass(frozen=True, slots=True)
 class CalendarInvitationAttendeeRequest:
     address: str
@@ -156,6 +165,27 @@ class CalendarSchedulingProvider(Protocol):
     ) -> CalendarEventRef:
         ...
 
+    def promote_event(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        user_id: str | None,
+        event_id: str,
+        request: CalendarEventRequest,
+    ) -> CalendarEventRef:
+        ...
+
+    def release_event(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        user_id: str | None,
+        event_id: str,
+    ) -> CalendarEventReleaseRef:
+        ...
+
 
 @runtime_checkable
 class CalendarOutboxProvider(Protocol):

src/govoplan_core/core/campaigns.py

@@ -3,14 +3,29 @@ from __future__ import annotations
 from collections.abc import Callable, Iterable, Mapping
 from dataclasses import dataclass, field
 from datetime import datetime
-from typing import Protocol, runtime_checkable
+from typing import Literal, Protocol, runtime_checkable
 
 
 CAPABILITY_CAMPAIGNS_MAIL_POLICY_CONTEXT = "campaigns.mailPolicyContext"
 CAPABILITY_CAMPAIGNS_ACCESS = "campaigns.access"
 CAPABILITY_CAMPAIGNS_POLICY_CONTEXT = "campaigns.policyContext"
 CAPABILITY_CAMPAIGNS_DELIVERY_TASKS = "campaigns.deliveryTasks"
+CAPABILITY_CAMPAIGNS_SCHEDULES = "campaigns.schedules"
 CAPABILITY_CAMPAIGNS_RETENTION = "campaigns.retention"
+CAPABILITY_CAMPAIGNS_WORK_ORCHESTRATION = "campaigns.workOrchestration"
+
+CampaignWorkAssigneeKind = Literal[
+    "account",
+    "group",
+    "organization_function",
+]
+CampaignWorkHandoffStatus = Literal[
+    "open",
+    "in_progress",
+    "completed",
+    "rejected",
+    "cancelled",
+]
 
 
 @dataclass(frozen=True, slots=True)
@@ -31,6 +46,88 @@ class CampaignPolicyContext:
     settings: Mapping[str, object] = field(default_factory=dict)
 
 
+@dataclass(frozen=True, slots=True)
+class CampaignWorkHandoffRequest:
+    """Typed request used by Workflow to open accountable Campaign work."""
+
+    tenant_id: str
+    idempotency_key: str
+    purpose: str
+    assignee_kind: CampaignWorkAssigneeKind
+    assignee_id: str
+    campaign_id: str | None = None
+    create_external_id: str | None = None
+    create_name: str | None = None
+    create_description: str | None = None
+    expected_campaign_revision: int | None = None
+    due_at: datetime | None = None
+    mirror_to_tasks: bool = True
+    correlation_id: str | None = None
+    workflow_instance_id: str | None = None
+    workflow_step_id: str | None = None
+
+    def __post_init__(self) -> None:
+        for value, label in (
+            (self.tenant_id, "Campaign hand-off tenant"),
+            (self.idempotency_key, "Campaign hand-off idempotency key"),
+            (self.purpose, "Campaign hand-off purpose"),
+            (self.assignee_id, "Campaign hand-off assignee"),
+        ):
+            if not value.strip():
+                raise ValueError(f"{label} is required")
+        references_existing = bool(self.campaign_id and self.campaign_id.strip())
+        creates_new = bool(
+            self.create_external_id
+            and self.create_external_id.strip()
+            and self.create_name
+            and self.create_name.strip()
+        )
+        if references_existing == creates_new:
+            raise ValueError(
+                "Campaign hand-offs must either reference one campaign or "
+                "declare one new campaign."
+            )
+        if self.expected_campaign_revision is not None and (
+            self.expected_campaign_revision < 1
+        ):
+            raise ValueError("Expected Campaign revisions start at one")
+        if self.due_at is not None and self.due_at.tzinfo is None:
+            raise ValueError("Campaign hand-off due dates require a timezone")
+
+
+@dataclass(frozen=True, slots=True)
+class CampaignWorkHandoffRef:
+    """Stable, revision-bearing reference returned to the Workflow instance."""
+
+    tenant_id: str
+    campaign_id: str
+    campaign_version_id: str
+    campaign_revision: int
+    assignment_id: str
+    assignment_revision: int
+    status: CampaignWorkHandoffStatus
+    action_url: str
+    campaign_ref: str
+    assignment_ref: str
+    event_type: str = "campaign.work.changed"
+    replayed: bool = False
+    optional_capabilities: Mapping[str, bool] = field(default_factory=dict)
+    provenance: Mapping[str, object] = field(default_factory=dict)
+
+
+@dataclass(frozen=True, slots=True)
+class CampaignWorkHandoffInspection:
+    """Current authorization and revision check before Workflow continuation."""
+
+    allowed: bool
+    status: CampaignWorkHandoffStatus | None = None
+    assignment_revision: int | None = None
+    action_url: str | None = None
+    assignment_ref: str | None = None
+    reason: str | None = None
+    provenance: Mapping[str, object] = field(default_factory=dict)
+
+
 @runtime_checkable
 class CampaignMailPolicyContextProvider(Protocol):
     def get_campaign_mail_policy_context(
@@ -105,6 +202,21 @@ class CampaignDeliveryTaskProvider(Protocol):
         ...
 
 
+@runtime_checkable
+class CampaignScheduleProvider(Protocol):
+    """Durable boundary for due manual drafts and governed autonomous occurrences."""
+
+    def dispatch_due(
+        self,
+        session: object,
+        *,
+        tenant_id: str | None = None,
+        now: datetime | None = None,
+        limit: int = 50,
+    ) -> Mapping[str, object]:
+        ...
+
+
 @runtime_checkable
 class CampaignRetentionProvider(Protocol):
     def apply_retention(
@@ -116,3 +228,45 @@ class CampaignRetentionProvider(Protocol):
         policy_for_campaign_id: Callable[[str | None], object],
     ) -> Mapping[str, Mapping[str, int]]:
         ...
+
+
+@runtime_checkable
+class CampaignWorkOrchestrationProvider(Protocol):
+    """Optional Campaign boundary for durable Workflow-owned hand-offs."""
+
+    def prepare_handoff(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: CampaignWorkHandoffRequest,
+    ) -> CampaignWorkHandoffRef:
+        ...
+
+    def inspect_handoff(
+        self,
+        session: object,
+        principal: object,
+        *,
+        tenant_id: str,
+        assignment_id: str,
+        expected_revision: int | None = None,
+    ) -> CampaignWorkHandoffInspection:
+        ...
+
+
+def campaign_work_orchestration_provider(
+    registry: object | None,
+) -> CampaignWorkOrchestrationProvider | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(CAPABILITY_CAMPAIGNS_WORK_ORCHESTRATION)
+    ):
+        return None
+    capability = registry.capability(CAPABILITY_CAMPAIGNS_WORK_ORCHESTRATION)
+    return (
+        capability
+        if isinstance(capability, CampaignWorkOrchestrationProvider)
+        else None
+    )

src/govoplan_core/core/concurrency.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import copy
 import hashlib
+import heapq
 import json
 from dataclasses import dataclass, field
 from typing import Any, Callable, Iterable, Mapping, Sequence
@@ -436,10 +437,33 @@ def _merge_list(
             result_by_id[identity] = merged.value
 
     order_source = local_order if local_reordered and not current_reordered else current_order
-    merged_order = [identity for identity in order_source if identity in result_by_id]
-    for identity in identities:
-        if identity in result_by_id and identity not in merged_order:
+    secondary_order = current_order if order_source is local_order else local_order
+    # Keep the chosen side's order and both sides' insertion anchors. Appending
+    # missing IDs would silently relocate an insertion during a disjoint edit.
+    # An incompatible reorder/insertion cycle is an explicit conflict.
+    edges: dict[str, set[str]] = {identity: set() for identity in result_by_id}
+    incoming = dict.fromkeys(result_by_id, 0)
+    for order, insertions_only in ((order_source, False), (secondary_order, True)):
+        selected = [identity for identity in order if identity in result_by_id]
+        for left, right in zip(selected, selected[1:]):
+            if insertions_only and left in base_by_id and right in base_by_id:
+                continue
+            if right not in edges[left]:
+                edges[left].add(right)
+                incoming[right] += 1
+    priority = {identity: index for index, identity in enumerate(identities)}
+    ready = [(priority[identity], identity) for identity, count in incoming.items() if count == 0]
+    heapq.heapify(ready)
+    merged_order: list[str] = []
+    while ready:
+        _, identity = heapq.heappop(ready)
         merged_order.append(identity)
+        for following in edges[identity]:
+            incoming[following] -= 1
+            if incoming[following] == 0:
+                heapq.heappush(ready, (priority[following], following))
+    if len(merged_order) != len(result_by_id):
+        return _conflict(path, "collection_reorder", base, local, current)
     return ThreeWayMergeResult(
         value=[result_by_id[identity] for identity in merged_order],
         conflicts=conflicts,

src/govoplan_core/core/configuration_control.py

@@ -371,6 +371,24 @@ def _approval_count(request: dict[str, Any]) -> int:
 
 
 def _sanitize_value(key: str, value: object) -> object:
+    if key == "campaign_archive_encryption_policy":
+        if not isinstance(value, dict):
+            return "<redacted>"
+        # These are format/channel names, never passwords. Preserve only the
+        # exact public enum lists so rollback history remains useful without
+        # exempting arbitrary password-named fields from secret redaction.
+        allowed_values = {
+            "allowed_password_encryption_methods": frozenset({"aes", "zip_standard"}),
+            "allowed_password_delivery_channels": frozenset({"separate_mail", "sms", "letter", "phone", "in_person"}),
+        }
+        return {
+            name: list(items)
+            if name in allowed_values
+            and isinstance(items, list)
+            and all(isinstance(item, str) and item in allowed_values[name] for item in items)
+            else "<redacted>"
+            for name, items in value.items()
+        }
     field = classify_configuration_field(key)
     if field is not None and field.secret_handling in {"reference_only", "env_only"}:
         return _redact_secrets(value)

src/govoplan_core/core/configuration_packages.py

@@ -3,6 +3,8 @@ from __future__ import annotations
 import base64
 from collections.abc import Mapping, Sequence
 from dataclasses import dataclass, field
+from datetime import UTC, datetime
+from importlib.metadata import PackageNotFoundError, version as package_version
 from pathlib import Path
 import json
 import os
@@ -28,6 +30,9 @@ from govoplan_core.core.external_references import (
     SourceAuthorityMode,
     integration_maturity_rank,
 )
+from govoplan_core.core.infrastructure_capabilities import (
+    InfrastructureCapabilityReceipt,
+)
 from govoplan_core.security.http_fetch import fetch_http_text
 
 
@@ -35,6 +40,12 @@ CONFIGURATION_PROVIDER_CAPABILITY = "configuration.provider"
 
 DiagnosticSeverity = Literal["blocker", "warning", "info"]
 PlanAction = Literal["create", "update", "bind", "skip", "blocked", "noop"]
+ConfigurationRollbackStatus = Literal[
+    "blocked_before_apply",
+    "not_required",
+    "database_restore_required",
+    "partial_apply_requires_recovery",
+]
 ConfigurationPackageClass = Literal[
     "reference",
     "product",
@@ -441,6 +452,9 @@ class ConfigurationPreflightContext:
         default_factory=dict
     )
     dry_run: bool = True
+    operator_scopes: frozenset[str] = frozenset()
+    infrastructure_receipt: InfrastructureCapabilityReceipt | None = None
+    infrastructure_receipt_error: str | None = None
 
 
 @dataclass(frozen=True, slots=True)
@@ -455,6 +469,21 @@ class ConfigurationApplyResult:
     diagnostics: tuple[ConfigurationDiagnostic, ...] = ()
     created_refs: Mapping[str, str] = field(default_factory=dict)
     updated_refs: Mapping[str, str] = field(default_factory=dict)
+    rollback: "ConfigurationRollbackState | None" = None
+
+
+@dataclass(frozen=True, slots=True)
+class ConfigurationRollbackState:
+    status: ConfigurationRollbackStatus
+    summary: str
+    recovery_action: str | None = None
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "status": self.status,
+            "summary": self.summary,
+            "recovery_action": self.recovery_action,
+        }
 
 
 @dataclass(frozen=True, slots=True)
@@ -465,11 +494,40 @@ class ConfigurationExportSelection:
     object_refs: tuple[str, ...] = ()
 
 
+@dataclass(frozen=True, slots=True)
+class ConfigurationExportProvenance:
+    exported_at: str
+    source_core_version: str
+    module_versions: Mapping[str, str]
+    tenant_id: str | None
+    exporter_id: str | None
+    scopes: tuple[str, ...] = ()
+    module_ids: tuple[str, ...] = ()
+    object_refs: tuple[str, ...] = ()
+    redacted_secret_keys: tuple[str, ...] = ()
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "exported_at": self.exported_at,
+            "source_core_version": self.source_core_version,
+            "module_versions": dict(self.module_versions),
+            "tenant_id": self.tenant_id,
+            "exporter_id": self.exporter_id,
+            "selection": {
+                "scopes": list(self.scopes),
+                "module_ids": list(self.module_ids),
+                "object_refs": list(self.object_refs),
+            },
+            "redacted_secret_keys": list(self.redacted_secret_keys),
+        }
+
+
 @dataclass(frozen=True, slots=True)
 class ConfigurationExportResult:
     fragments: tuple[ConfigurationPackageFragment, ...] = ()
     data_requirements: tuple[ConfigurationRequiredData, ...] = ()
     diagnostics: tuple[ConfigurationDiagnostic, ...] = ()
+    provenance: ConfigurationExportProvenance | None = None
 
 
 @runtime_checkable
@@ -502,6 +560,7 @@ def dry_run_configuration_package(
     diagnostics: list[ConfigurationDiagnostic] = []
     required_data: list[ConfigurationRequiredData] = []
     plan: list[ConfigurationPlanItem] = []
+    declared_data: dict[str, ConfigurationRequiredData] = {}
 
     diagnostics.extend(_module_requirement_diagnostics(manifest, context))
     diagnostics.extend(_capability_requirement_diagnostics(manifest, context))
@@ -509,6 +568,7 @@ def dry_run_configuration_package(
     for item in manifest.data_requirements:
         requirement = ConfigurationRequiredData.from_mapping(item)
         required_data.append(requirement)
+        declared_data[requirement.key] = requirement
         if requirement.required and requirement.key not in context.supplied_data:
             diagnostics.append(ConfigurationDiagnostic(
                 severity="blocker",
@@ -519,6 +579,25 @@ def dry_run_configuration_package(
             ))
 
     for fragment in manifest.fragments:
+        data_ref_diagnostics = _fragment_data_reference_diagnostics(
+            fragment,
+            declared_data=declared_data,
+            supplied_data=context.supplied_data,
+        )
+        if data_ref_diagnostics:
+            diagnostics.extend(data_ref_diagnostics)
+            plan.append(ConfigurationPlanItem(
+                action="blocked",
+                module_id=fragment.module_id,
+                fragment_type=fragment.fragment_type,
+                fragment_id=fragment.fragment_id,
+                summary="Fragment needs declared deployment data before provider preflight.",
+            ))
+            continue
+        resolved_fragment = _resolve_fragment_data_references(
+            fragment,
+            context.supplied_data,
+        )
         provider = provider_map.get(fragment.module_id)
         if provider is None:
             diagnostics.append(ConfigurationDiagnostic(
@@ -544,7 +623,7 @@ def dry_run_configuration_package(
             plan.append(ConfigurationPlanItem(action="blocked", module_id=fragment.module_id, fragment_type=fragment.fragment_type, fragment_id=fragment.fragment_id, summary="Fragment type is unsupported."))
             continue
         try:
-            result = provider.preflight(fragment, context)
+            result = provider.preflight(resolved_fragment, context)
         except Exception as exc:
             diagnostics.append(ConfigurationDiagnostic(
                 severity="blocker",
@@ -586,29 +665,54 @@ def apply_configuration_package(
     apply_context = ConfigurationPreflightContext(
         tenant_id=context.tenant_id,
         operator_user_id=context.operator_user_id,
+        operator_scopes=context.operator_scopes,
         supplied_data=supplied_data if supplied_data is not None else context.supplied_data,
         installed_modules=context.installed_modules,
         capabilities=context.capabilities,
         external_provider_declarations=context.external_provider_declarations,
         external_provider_states=context.external_provider_states,
+        infrastructure_receipt=context.infrastructure_receipt,
+        infrastructure_receipt_error=context.infrastructure_receipt_error,
         dry_run=False,
     )
     preflight = dry_run_configuration_package(manifest, providers, apply_context)
     blockers = [item for item in preflight.diagnostics if item.severity == "blocker"]
     if blockers:
-        return ConfigurationApplyResult(diagnostics=tuple(blockers))
+        return ConfigurationApplyResult(
+            diagnostics=tuple(blockers),
+            rollback=ConfigurationRollbackState(
+                status="blocked_before_apply",
+                summary="No provider changes were attempted because package preflight is blocked.",
+            ),
+        )
     provider_map = _configuration_provider_map(providers)
     diagnostics: list[ConfigurationDiagnostic] = list(preflight.diagnostics)
     created_refs: dict[str, str] = {}
     updated_refs: dict[str, str] = {}
+    stopped_after_blocker = False
     for fragment in manifest.fragments:
         provider = provider_map[fragment.module_id]
+        resolved_fragment = _resolve_fragment_data_references(
+            fragment,
+            apply_context.supplied_data,
+        )
         try:
-            result = provider.apply(fragment, apply_context.supplied_data, apply_context)
+            result = provider.apply(
+                resolved_fragment,
+                apply_context.supplied_data,
+                apply_context,
+            )
             diagnostics.extend(result.diagnostics)
             created_refs.update(result.created_refs)
             updated_refs.update(result.updated_refs)
-            diagnostics.extend(provider.health(result, apply_context))
+            health_diagnostics = provider.health(result, apply_context)
+            diagnostics.extend(health_diagnostics)
+            if any(
+                item.severity == "blocker"
+                for item in (*result.diagnostics, *health_diagnostics)
+            ):
+                stopped_after_blocker = True
+                break
         except Exception as exc:
             diagnostics.append(ConfigurationDiagnostic(
                 severity="blocker",
@@ -618,10 +722,36 @@ def apply_configuration_package(
                 object_ref=fragment.fragment_id or fragment.fragment_type,
                 resolution="Stop the import, keep previous configuration, and inspect provider logs.",
             ))
+            stopped_after_blocker = True
+            break
+    changed = bool(created_refs or updated_refs)
+    if stopped_after_blocker and changed:
+        rollback = ConfigurationRollbackState(
+            status="partial_apply_requires_recovery",
+            summary="At least one provider committed changes before a later provider blocked the package.",
+            recovery_action="Restore the reviewed pre-apply database snapshot or use module-owned compensation where explicitly supported.",
+        )
+    elif stopped_after_blocker:
+        rollback = ConfigurationRollbackState(
+            status="blocked_before_apply",
+            summary="The first provider blocked before any configuration reference was created or updated.",
+        )
+    elif changed:
+        rollback = ConfigurationRollbackState(
+            status="database_restore_required",
+            summary="The package changed provider-owned configuration; generic cross-module compensation is not available.",
+            recovery_action="Retain the pre-apply database snapshot until verification is complete; restore it if the package must be rolled back.",
+        )
+    else:
+        rollback = ConfigurationRollbackState(
+            status="not_required",
+            summary="All package fragments were no-ops, so no rollback action is required.",
+        )
     return ConfigurationApplyResult(
         diagnostics=tuple(_dedupe_diagnostics(diagnostics)),
         created_refs=created_refs,
         updated_refs=updated_refs,
+        rollback=rollback,
     )
 
 
@@ -660,10 +790,29 @@ def export_configuration_package(
         fragments.extend(result.fragments)
         data_requirements.extend(result.data_requirements)
         diagnostics.extend(result.diagnostics)
+    deduped_required_data = tuple(_dedupe_required_data(data_requirements))
+    provenance = ConfigurationExportProvenance(
+        exported_at=datetime.now(UTC).isoformat(),
+        source_core_version=_installed_core_version(),
+        module_versions={
+            module_id: context.installed_modules[module_id]
+            for module_id in sorted(set(module_ids))
+            if module_id in context.installed_modules
+        },
+        tenant_id=selection.tenant_id,
+        exporter_id=context.operator_user_id,
+        scopes=selection.scopes,
+        module_ids=tuple(module_ids),
+        object_refs=selection.object_refs,
+        redacted_secret_keys=tuple(
+            sorted(item.key for item in deduped_required_data if item.secret)
+        ),
+    )
     return ConfigurationExportResult(
         fragments=tuple(fragments),
-        data_requirements=tuple(_dedupe_required_data(data_requirements)),
+        data_requirements=deduped_required_data,
         diagnostics=tuple(_dedupe_diagnostics(diagnostics)),
+        provenance=provenance,
     )
 
 
@@ -1274,6 +1423,103 @@ def _dedupe_required_data(items: Sequence[ConfigurationRequiredData]) -> list[Co
     return result
 
 
+def _fragment_data_reference_diagnostics(
+    fragment: ConfigurationPackageFragment,
+    *,
+    declared_data: Mapping[str, ConfigurationRequiredData],
+    supplied_data: Mapping[str, Any],
+) -> list[ConfigurationDiagnostic]:
+    references: set[str] = set()
+    invalid = _collect_fragment_data_references(fragment.payload, references)
+    diagnostics: list[ConfigurationDiagnostic] = []
+    object_ref = fragment.fragment_id or fragment.fragment_type
+    if invalid:
+        diagnostics.append(ConfigurationDiagnostic(
+            severity="blocker",
+            code="fragment_data_reference_invalid",
+            message="Configuration fragment data references must be objects containing only a non-empty $data key.",
+            module_id=fragment.module_id,
+            object_ref=object_ref,
+            resolution="Replace malformed references with {\"$data\": \"declared_requirement_key\"}.",
+        ))
+    for key in sorted(references - set(declared_data)):
+        diagnostics.append(ConfigurationDiagnostic(
+            severity="blocker",
+            code="fragment_data_reference_undeclared",
+            message=f"Configuration fragment references undeclared operator data {key!r}.",
+            module_id=fragment.module_id,
+            object_ref=key,
+            resolution="Declare the key in package data_requirements before using it in a fragment.",
+        ))
+    for key in sorted(references & set(declared_data)):
+        if key in supplied_data:
+            continue
+        diagnostics.append(ConfigurationDiagnostic(
+            severity="blocker",
+            code="fragment_data_reference_missing",
+            message=f"Configuration fragment needs operator data {declared_data[key].label!r} before provider preflight.",
+            module_id=fragment.module_id,
+            object_ref=key,
+            resolution="Provide the value in the generated configuration package form.",
+        ))
+    return diagnostics
+
+
+def _collect_fragment_data_references(value: object, references: set[str]) -> bool:
+    invalid = False
+    if isinstance(value, Mapping):
+        if "$data" in value:
+            key = value.get("$data")
+            if len(value) != 1 or not isinstance(key, str) or not key.strip():
+                return True
+            references.add(key.strip())
+            return False
+        for item in value.values():
+            invalid = _collect_fragment_data_references(item, references) or invalid
+    elif isinstance(value, Sequence) and not isinstance(value, (str, bytes)):
+        for item in value:
+            invalid = _collect_fragment_data_references(item, references) or invalid
+    return invalid
+
+
+def _resolve_fragment_data_references(
+    fragment: ConfigurationPackageFragment,
+    supplied_data: Mapping[str, Any],
+) -> ConfigurationPackageFragment:
+    payload = _resolve_data_reference_value(fragment.payload, supplied_data)
+    if not isinstance(payload, Mapping):
+        raise ValueError("Resolved configuration fragment payload must remain an object.")
+    return ConfigurationPackageFragment(
+        module_id=fragment.module_id,
+        fragment_type=fragment.fragment_type,
+        fragment_id=fragment.fragment_id,
+        payload=payload,
+    )
+
+
+def _resolve_data_reference_value(value: object, supplied_data: Mapping[str, Any]) -> object:
+    if isinstance(value, Mapping):
+        if set(value) == {"$data"}:
+            key = value.get("$data")
+            if not isinstance(key, str) or key not in supplied_data:
+                raise ValueError("Configuration fragment contains an unresolved $data reference.")
+            return supplied_data[key]
+        return {
+            str(key): _resolve_data_reference_value(item, supplied_data)
+            for key, item in value.items()
+        }
+    if isinstance(value, Sequence) and not isinstance(value, (str, bytes)):
+        return [_resolve_data_reference_value(item, supplied_data) for item in value]
+    return value
+
+
+def _installed_core_version() -> str:
+    try:
+        return package_version("govoplan-core")
+    except PackageNotFoundError:
+        return "workspace"
+
+
 def _catalog_source(path: Path | str | None) -> Path | str | None:
     if path is not None:
         return path if isinstance(path, str) and _is_http_url(path) else Path(path).expanduser()

src/govoplan_core/core/configuration_safety.py

@@ -107,6 +107,20 @@ class _ConfigurationChangeSafetyState:
 
 
 _CONFIGURATION_FIELD_SAFETY: tuple[ConfigurationFieldSafety, ...] = (
+    ConfigurationFieldSafety(
+        key="campaign_delivery_policy.system", label="System Campaign synchronous delivery limit",
+        owner_module="campaigns", scope="system", storage="system_settings", ui_managed=True,
+        risk="medium", required_scopes=("system:settings:write",),
+        audit_event="campaign.delivery_policy_updated", rollback_history_required=True,
+        notes="A bounded 0–500 recipient-job maximum for one interactive Send now request. Explicit deployment ceilings remain authoritative; saving never delivers mail or changes review evidence.",
+    ),
+    ConfigurationFieldSafety(
+        key="campaign_delivery_policy.tenant", label="Tenant Campaign synchronous delivery limit",
+        owner_module="campaigns", scope="tenant", storage="tenant_settings", ui_managed=True,
+        risk="medium", required_scopes=("admin:policies:write",),
+        audit_event="campaign.delivery_policy_updated", rollback_history_required=True,
+        notes="Tenant policy may only narrow the inherited system/deployment recipient-job maximum; clearing an override restores inheritance. Changes retain before/after history.",
+    ),
     ConfigurationFieldSafety(
         key="module_management.desired_enabled",
         label="Enabled modules",
@@ -171,6 +185,21 @@ _CONFIGURATION_FIELD_SAFETY: tuple[ConfigurationFieldSafety, ...] = (
         rollback_history_required=True,
         notes="Maintenance mode controls platform availability and gates dangerous operations.",
     ),
+    ConfigurationFieldSafety(
+        key="campaign_archive_encryption_policy",
+        label="Campaign archive encryption policy",
+        owner_module="policy",
+        scope="system",
+        storage="policy_overrides",
+        ui_managed=True,
+        risk="high",
+        required_scopes=("system:settings:write", "admin:policies:write"),
+        validation_required=True,
+        policy_explanation_required=True,
+        audit_event="campaign_archive_encryption_policy.updated",
+        rollback_history_required=True,
+        notes="Explicit system ceiling for Campaign archive methods and separate password-delivery channels. Policy validates allowed values and retains before/after history; lower scopes may only narrow. Legacy use additionally requires the dedicated Campaign permission and reasoned weak-encryption acknowledgement, so saving policy alone never enables or sends an archive.",
+    ),
     ConfigurationFieldSafety(
         key="privacy_retention_policy",
         label="Privacy retention policy",

src/govoplan_core/core/datasources.py

@@ -5,6 +5,7 @@ from dataclasses import dataclass, field
 from datetime import datetime
 from typing import Literal, Protocol, runtime_checkable
 
+from govoplan_core.core.access import PrincipalRef
 from govoplan_core.core.external_references import (
     SOURCE_AUTHORITY_MODES,
     SourceAuthorityMode,
@@ -13,6 +14,7 @@ from govoplan_core.core.tabular_sources import (
     DEFAULT_PREVIEW_BYTES,
     DEFAULT_PREVIEW_TIMEOUT_MS,
     TabularPreviewDiagnostic,
+    TabularCsvSource,
     TabularPushdown,
     TabularSourceHealth,
     TabularSourceMode,
@@ -23,6 +25,8 @@ CAPABILITY_DATASOURCE_CATALOGUE = "datasources.catalogue"
 CAPABILITY_DATASOURCE_LIFECYCLE = "datasources.lifecycle"
 CAPABILITY_DATASOURCE_PUBLICATION = "datasources.publication"
 CAPABILITY_DATASOURCE_ORIGINS = "connectors.datasourceOrigins"
+CAPABILITY_DATASOURCE_ARTIFACT_BACKENDS = "datasources.artifactBackends"
+CAPABILITY_POLICY_DATASOURCE_VISIBILITY = "policy.datasourceVisibility"
 
 DatasourceMode = Literal["live", "cached", "static"]
 DatasourceKind = Literal[
@@ -37,6 +41,12 @@ DatasourceKind = Literal[
 ]
 DatasourceShape = Literal["tabular", "document", "binary", "directory", "stream"]
 DatasourceConsistency = Literal["current", "live", "frozen"]
+DatasourceVisibilityAction = Literal["discover", "read"]
+DatasourcePublicationStatus = Literal[
+    "published",
+    "published_with_warnings",
+    "review_required",
+]
 
 
 class DatasourceError(ValueError):
@@ -64,6 +74,7 @@ class DatasourceField:
     name: str
     data_type: str
     nullable: bool = True
+    classification: str = "internal"
 
 
 @dataclass(frozen=True, slots=True)
@@ -84,11 +95,15 @@ class DatasourceGovernance:
     classification: str = "internal"
     privacy_profile_ref: str | None = None
     retention_policy_ref: str | None = None
+    access_policy_ref: str | None = None
+    visibility_policy: Mapping[str, object] = field(default_factory=dict)
     hold_refs: tuple[str, ...] = ()
     publication_state: str = "draft"
     transfer_agreement_ref: str | None = None
     freshness_policy: Mapping[str, object] = field(default_factory=dict)
     quality_policy: Mapping[str, object] = field(default_factory=dict)
+    approval_policy: Mapping[str, object] = field(default_factory=dict)
+    retention_policy: Mapping[str, object] = field(default_factory=dict)
     known_limits: tuple[str, ...] = ()
     correction_procedure_ref: str | None = None
     affected_refs: tuple[str, ...] = ()
@@ -154,6 +169,12 @@ class DatasourceGovernance:
             retention_policy_ref=_optional_governance_text(
                 source.get("retention_policy_ref")
             ),
+            access_policy_ref=_optional_governance_text(
+                source.get("access_policy_ref")
+            ),
+            visibility_policy=_governance_mapping(
+                source.get("visibility_policy")
+            ),
             hold_refs=_governance_texts(source.get("hold_refs")),
             publication_state=str(source.get("publication_state") or "draft"),
             transfer_agreement_ref=_optional_governance_text(
@@ -161,6 +182,8 @@ class DatasourceGovernance:
             ),
             freshness_policy=_governance_mapping(source.get("freshness_policy")),
             quality_policy=_governance_mapping(source.get("quality_policy")),
+            approval_policy=_governance_mapping(source.get("approval_policy")),
+            retention_policy=_governance_mapping(source.get("retention_policy")),
             known_limits=_governance_texts(source.get("known_limits")),
             correction_procedure_ref=_optional_governance_text(
                 source.get("correction_procedure_ref")
@@ -185,11 +208,15 @@ class DatasourceGovernance:
             "classification": self.classification,
             "privacy_profile_ref": self.privacy_profile_ref,
             "retention_policy_ref": self.retention_policy_ref,
+            "access_policy_ref": self.access_policy_ref,
+            "visibility_policy": dict(self.visibility_policy),
             "hold_refs": list(self.hold_refs),
             "publication_state": self.publication_state,
             "transfer_agreement_ref": self.transfer_agreement_ref,
             "freshness_policy": dict(self.freshness_policy),
             "quality_policy": dict(self.quality_policy),
+            "approval_policy": dict(self.approval_policy),
+            "retention_policy": dict(self.retention_policy),
             "known_limits": list(self.known_limits),
             "correction_procedure_ref": self.correction_procedure_ref,
             "affected_refs": list(self.affected_refs),
@@ -197,6 +224,39 @@ class DatasourceGovernance:
         }
 
 
+@dataclass(frozen=True, slots=True)
+class DatasourceVisibilityPolicyRequest:
+    tenant_id: str
+    datasource_ref: str
+    principal: PrincipalRef
+    action: DatasourceVisibilityAction
+    classification: str = "internal"
+    policy_ref: str | None = None
+    consistency: DatasourceConsistency = "current"
+    materialization_ref: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class DatasourceVisibilityPolicyDecision:
+    allowed: bool
+    reason: str | None = None
+    policies: tuple[Mapping[str, object], ...] = ()
+    decision_ref: str | None = None
+    provenance: Mapping[str, object] = field(default_factory=dict)
+
+
+@runtime_checkable
+class DatasourceVisibilityPolicyProvider(Protocol):
+    """Optionally tighten Datasources-owned local visibility policy."""
+
+    def decide_datasource_visibility(
+        self,
+        session: object,
+        *,
+        request: DatasourceVisibilityPolicyRequest,
+    ) -> DatasourceVisibilityPolicyDecision: ...
+
+
 @dataclass(frozen=True, slots=True)
 class DatasourceDescriptor:
     ref: str
@@ -236,6 +296,8 @@ class DatasourceMaterialization:
     frozen_label: str | None = None
     source_timestamp: datetime | None = None
     created_at: datetime | None = None
+    disposed_at: datetime | None = None
+    disposition: Mapping[str, object] = field(default_factory=dict)
     provenance: Mapping[str, object] = field(default_factory=dict)
     metadata: Mapping[str, object] = field(default_factory=dict)
     governance: DatasourceGovernance = field(default_factory=DatasourceGovernance)
@@ -256,6 +318,7 @@ class DatasourceStage:
     row_count: int | None = None
     byte_count: int | None = None
     validation: Mapping[str, object] = field(default_factory=dict)
+    approval: Mapping[str, object] = field(default_factory=dict)
     created_at: datetime | None = None
     promoted_at: datetime | None = None
     promoted_materialization_ref: str | None = None
@@ -307,6 +370,67 @@ class DatasourceStageInput:
     provenance: Mapping[str, object] = field(default_factory=dict)
     metadata: Mapping[str, object] = field(default_factory=dict)
     governance: DatasourceGovernance | None = None
+    csv_source: TabularCsvSource | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class DatasourceArtifactReference:
+    """Immutable provider-neutral reference to a durable tabular payload.
+
+    The producer owns creation of the payload. Datasources pins its locator,
+    checksum and declared shape without importing the artifact-owning module;
+    a configured payload backend verifies integrity and provides bounded reads.
+    """
+
+    backend: str
+    locator: str
+    checksum: str
+    row_count: int
+    byte_count: int
+    schema: tuple[DatasourceField, ...]
+    fingerprint: str
+    media_type: str = "application/x-ndjson"
+    checkpoint: Mapping[str, object] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
+    validation: Mapping[str, object] = field(default_factory=dict)
+
+
+@runtime_checkable
+class DatasourceArtifactBackend(Protocol):
+    """Storage-module boundary for immutable artifact-backed tabular data."""
+
+    backend: str
+
+    def verify(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        artifact: DatasourceArtifactReference,
+    ) -> None: ...
+
+    def read_rows(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        artifact: DatasourceArtifactReference,
+        offset: int,
+        limit: int,
+    ) -> Sequence[Mapping[str, object]]: ...
+
+    def delete(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        artifact: DatasourceArtifactReference,
+    ) -> None: ...
+
+
+@runtime_checkable
+class DatasourceArtifactBackendProvider(Protocol):
+    def artifact_backends(self) -> Sequence[DatasourceArtifactBackend]: ...
 
 
 @dataclass(frozen=True, slots=True)
@@ -314,7 +438,8 @@ class DatasourcePublicationRequest:
     producer_module: str
     producer_run_ref: str
     idempotency_key: str
-    rows: tuple[Mapping[str, object], ...]
+    rows: tuple[Mapping[str, object], ...] | None = None
+    artifact: DatasourceArtifactReference | None = None
     target_datasource_ref: str | None = None
     name: str | None = None
     source_name: str | None = None
@@ -331,7 +456,7 @@ class DatasourcePublicationRequest:
 @dataclass(frozen=True, slots=True)
 class DatasourcePublicationResult:
     ref: str
-    status: str
+    status: DatasourcePublicationStatus
     datasource: DatasourceDescriptor
     materialization: DatasourceMaterialization
     replayed: bool = False
@@ -564,6 +689,17 @@ def datasource_catalogue(registry: object | None) -> DatasourceCatalogueProvider
     return capability if isinstance(capability, DatasourceCatalogueProvider) else None
 
 
+def datasource_artifact_backend_provider(
+    registry: object | None,
+) -> DatasourceArtifactBackendProvider | None:
+    capability = _capability(registry, CAPABILITY_DATASOURCE_ARTIFACT_BACKENDS)
+    return (
+        capability
+        if isinstance(capability, DatasourceArtifactBackendProvider)
+        else None
+    )
+
+
 def datasource_lifecycle(registry: object | None) -> DatasourceLifecycleProvider | None:
     capability = _capability(registry, CAPABILITY_DATASOURCE_LIFECYCLE)
     return capability if isinstance(capability, DatasourceLifecycleProvider) else None
@@ -581,6 +717,13 @@ def datasource_origins(registry: object | None) -> DatasourceOriginProvider | No
     return capability if isinstance(capability, DatasourceOriginProvider) else None
 
 
+def datasource_visibility_policy_provider(
+    registry: object | None,
+) -> DatasourceVisibilityPolicyProvider | None:
+    capability = _capability(registry, CAPABILITY_POLICY_DATASOURCE_VISIBILITY)
+    return capability if isinstance(capability, DatasourceVisibilityPolicyProvider) else None
+
+
 def _capability(registry: object | None, name: str) -> object | None:
     if (
         registry is None
@@ -613,9 +756,15 @@ def _governance_mapping(value: object) -> Mapping[str, object]:
 
 __all__ = [
     "CAPABILITY_DATASOURCE_CATALOGUE",
+    "CAPABILITY_DATASOURCE_ARTIFACT_BACKENDS",
     "CAPABILITY_DATASOURCE_LIFECYCLE",
     "CAPABILITY_DATASOURCE_ORIGINS",
+    "CAPABILITY_DATASOURCE_PUBLICATION",
+    "CAPABILITY_POLICY_DATASOURCE_VISIBILITY",
     "DatasourceAccessError",
+    "DatasourceArtifactReference",
+    "DatasourceArtifactBackend",
+    "DatasourceArtifactBackendProvider",
     "DatasourceCatalogueProvider",
     "DatasourceConsistency",
     "DatasourceDescriptor",
@@ -631,6 +780,10 @@ __all__ = [
     "DatasourceOriginProvider",
     "DatasourceOriginReadRequest",
     "DatasourceOriginReadResult",
+    "DatasourcePublicationProvider",
+    "DatasourcePublicationRequest",
+    "DatasourcePublicationResult",
+    "DatasourcePublicationStatus",
     "DatasourceReadRequest",
     "DatasourceReadResult",
     "DatasourceShape",
@@ -638,7 +791,14 @@ __all__ = [
     "DatasourceStageInput",
     "DatasourceUnavailableError",
     "DatasourceValidationError",
+    "DatasourceVisibilityAction",
+    "DatasourceVisibilityPolicyDecision",
+    "DatasourceVisibilityPolicyProvider",
+    "DatasourceVisibilityPolicyRequest",
     "datasource_catalogue",
+    "datasource_artifact_backend_provider",
     "datasource_lifecycle",
     "datasource_origins",
+    "datasource_publication",
+    "datasource_visibility_policy_provider",
 ]

src/govoplan_core/core/dsar.py

@@ -0,0 +1,197 @@
+from __future__ import annotations
+
+from collections.abc import Mapping, Sequence
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Literal, Protocol, runtime_checkable
+
+
+DSAR_CAPABILITY_PREFIX = "privacy.dsar."
+
+DsarRequestKind = Literal["access", "erasure", "access_and_erasure"]
+DsarActionKind = Literal[
+    "delete",
+    "anonymize",
+    "revoke",
+    "detach",
+    "retain",
+    "manual_review",
+]
+
+
+@dataclass(frozen=True, slots=True)
+class DsarSubjectRef:
+    account_id: str | None = None
+    identity_id: str | None = None
+    membership_id: str | None = None
+    email: str | None = None
+    external_references: Mapping[str, str] = field(default_factory=dict)
+
+    def has_selector(self) -> bool:
+        return bool(
+            self.account_id
+            or self.identity_id
+            or self.membership_id
+            or self.email
+            or self.external_references
+        )
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "account_id": self.account_id,
+            "identity_id": self.identity_id,
+            "membership_id": self.membership_id,
+            "email": self.email,
+            "external_references": dict(self.external_references),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class DsarRecordRef:
+    provider_id: str
+    module_id: str
+    resource_type: str
+    resource_id: str
+    category: str
+    title: str
+    data: Mapping[str, object] = field(default_factory=dict)
+    observed_at: datetime | None = None
+    immutable_evidence: bool = False
+    retention_reason: str | None = None
+    source_path: str | None = None
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "provider_id": self.provider_id,
+            "module_id": self.module_id,
+            "resource_type": self.resource_type,
+            "resource_id": self.resource_id,
+            "category": self.category,
+            "title": self.title,
+            "data": dict(self.data),
+            "observed_at": self.observed_at.isoformat() if self.observed_at else None,
+            "immutable_evidence": self.immutable_evidence,
+            "retention_reason": self.retention_reason,
+            "source_path": self.source_path,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class DsarErasureActionRef:
+    action_id: str
+    provider_id: str
+    module_id: str
+    kind: DsarActionKind
+    resource_type: str
+    resource_id: str
+    title: str
+    rationale: str
+    executable: bool
+    irreversible: bool = False
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "action_id": self.action_id,
+            "provider_id": self.provider_id,
+            "module_id": self.module_id,
+            "kind": self.kind,
+            "resource_type": self.resource_type,
+            "resource_id": self.resource_id,
+            "title": self.title,
+            "rationale": self.rationale,
+            "executable": self.executable,
+            "irreversible": self.irreversible,
+            "metadata": dict(self.metadata),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class DsarExecutionResultRef:
+    action_id: str
+    status: Literal["executed", "unchanged", "failed", "blocked"]
+    summary: str
+    evidence: Mapping[str, object] = field(default_factory=dict)
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "action_id": self.action_id,
+            "status": self.status,
+            "summary": self.summary,
+            "evidence": dict(self.evidence),
+        }
+
+
+@runtime_checkable
+class DsarProvider(Protocol):
+    provider_id: str
+    module_id: str
+
+    def search_subject(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        subject: DsarSubjectRef,
+    ) -> Sequence[DsarRecordRef]: ...
+
+    def plan_erasure(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        subject: DsarSubjectRef,
+        records: Sequence[DsarRecordRef],
+    ) -> Sequence[DsarErasureActionRef]: ...
+
+    def execute_erasure(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        subject: DsarSubjectRef,
+        actions: Sequence[DsarErasureActionRef],
+        request_id: str,
+    ) -> Sequence[DsarExecutionResultRef]: ...
+
+
+def dsar_capability_name(module_id: str) -> str:
+    normalized = module_id.strip().casefold()
+    if not normalized or not normalized.replace("_", "").isalnum():
+        raise ValueError("DSAR module id must be an identifier.")
+    return f"{DSAR_CAPABILITY_PREFIX}{normalized}"
+
+
+def dsar_provider_names(registry: object | None) -> tuple[str, ...]:
+    if registry is None or not hasattr(registry, "capability_names"):
+        return ()
+    return tuple(
+        name
+        for name in registry.capability_names()
+        if name.startswith(DSAR_CAPABILITY_PREFIX)
+    )
+
+
+def dsar_provider(
+    registry: object,
+    capability_name: str,
+) -> DsarProvider:
+    provider = registry.require_capability(capability_name)
+    if not isinstance(provider, DsarProvider):
+        raise TypeError(f"{capability_name} does not implement DsarProvider")
+    return provider
+
+
+__all__ = [
+    "DSAR_CAPABILITY_PREFIX",
+    "DsarActionKind",
+    "DsarErasureActionRef",
+    "DsarExecutionResultRef",
+    "DsarProvider",
+    "DsarRecordRef",
+    "DsarRequestKind",
+    "DsarSubjectRef",
+    "dsar_capability_name",
+    "dsar_provider",
+    "dsar_provider_names",
+]

src/govoplan_core/core/files.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 from collections.abc import Mapping
 from dataclasses import dataclass, field
+from datetime import datetime
 from typing import Protocol, runtime_checkable
 
 from govoplan_core.core.access import ResourceAccessExplanationProvider
@@ -9,6 +10,49 @@ from govoplan_core.core.access import ResourceAccessExplanationProvider
 
 CAPABILITY_FILES_ACCESS = "files.access"
 CAPABILITY_FILES_ARTIFACT_STORE = "files.artifact_store"
+CAPABILITY_FILES_POSTBOX_REFERENCES = "files.postbox_references"
+CAPABILITY_FILES_TABULAR_CONTENT = "files.tabular_content"
+
+
+class ManagedTabularFileError(ValueError):
+    """Stable base error for exact-version managed tabular file access."""
+
+
+class ManagedTabularFileNotFoundError(ManagedTabularFileError):
+    pass
+
+
+class ManagedTabularFileAccessError(ManagedTabularFileError):
+    pass
+
+
+class ManagedTabularFileUnavailableError(ManagedTabularFileError):
+    pass
+
+
+class ManagedTabularFileValidationError(ManagedTabularFileError):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class ManagedTabularFile:
+    """Authorized metadata for one immutable managed file version."""
+
+    file_asset_id: str
+    file_version_id: str
+    filename: str
+    display_path: str
+    content_type: str | None
+    size_bytes: int
+    sha256: str
+    updated_at: datetime | None = None
+    current_version: bool = True
+
+
+@dataclass(frozen=True, slots=True)
+class ManagedTabularFileContent:
+    file: ManagedTabularFile
+    payload: bytes
 
 
 @dataclass(frozen=True, slots=True)
@@ -34,6 +78,30 @@ class ManagedArtifactRef:
     provenance: Mapping[str, object] = field(default_factory=dict)
 
 
+@dataclass(frozen=True, slots=True)
+class PostboxFileReferenceRequest:
+    reference_type: str
+    reference_id: str
+    postbox_id: str
+    message_id: str
+
+
+@dataclass(frozen=True, slots=True)
+class PostboxFileReferenceRef:
+    reference_type: str
+    reference_id: str
+    available: bool
+    reason_code: str
+    file_asset_id: str | None = None
+    file_version_id: str | None = None
+    filename: str | None = None
+    content_type: str | None = None
+    size_bytes: int | None = None
+    sha256: str | None = None
+    download_path: str | None = None
+    provenance: Mapping[str, object] = field(default_factory=dict)
+
+
 @runtime_checkable
 class FileAccessProvider(ResourceAccessExplanationProvider, Protocol):
     """Resource-level access explanation provider for Files-owned resources."""
@@ -50,3 +118,111 @@ class ManagedArtifactStore(Protocol):
         *,
         request: ManagedArtifactWriteRequest,
     ) -> ManagedArtifactRef: ...
+
+
+@runtime_checkable
+class PostboxFileReferenceProvider(Protocol):
+    """Resolve Files-owned references after Postbox and Files authorization."""
+
+    def resolve_postbox_references(
+        self,
+        session: object,
+        principal: object,
+        *,
+        tenant_id: str,
+        requests: tuple[PostboxFileReferenceRequest, ...],
+    ) -> tuple[PostboxFileReferenceRef, ...]: ...
+
+
+@runtime_checkable
+class ManagedTabularFileProvider(Protocol):
+    """List and open authorized CSV/XLSX content without exposing Files internals."""
+
+    def list_tabular_files(
+        self,
+        session: object,
+        principal: object,
+        *,
+        query: str = "",
+        limit: int = 100,
+    ) -> tuple[ManagedTabularFile, ...]: ...
+
+    def get_tabular_file(
+        self,
+        session: object,
+        principal: object,
+        *,
+        file_asset_id: str,
+        file_version_id: str | None = None,
+    ) -> ManagedTabularFile | None: ...
+
+    def read_tabular_file(
+        self,
+        session: object,
+        principal: object,
+        *,
+        file_asset_id: str,
+        file_version_id: str,
+        max_bytes: int,
+    ) -> ManagedTabularFileContent: ...
+
+
+def postbox_file_reference_provider(
+    registry: object | None,
+) -> PostboxFileReferenceProvider | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(CAPABILITY_FILES_POSTBOX_REFERENCES)
+    ):
+        return None
+    provider = registry.require_capability(CAPABILITY_FILES_POSTBOX_REFERENCES)
+    if not isinstance(provider, PostboxFileReferenceProvider):
+        raise TypeError(
+            "files.postbox_references provider does not implement "
+            "PostboxFileReferenceProvider"
+        )
+    return provider
+
+
+def managed_tabular_file_provider(
+    registry: object | None,
+) -> ManagedTabularFileProvider | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(CAPABILITY_FILES_TABULAR_CONTENT)
+    ):
+        return None
+    provider = registry.require_capability(CAPABILITY_FILES_TABULAR_CONTENT)
+    if not isinstance(provider, ManagedTabularFileProvider):
+        raise TypeError(
+            "files.tabular_content provider does not implement "
+            "ManagedTabularFileProvider"
+        )
+    return provider
+
+
+__all__ = [
+    "CAPABILITY_FILES_ACCESS",
+    "CAPABILITY_FILES_ARTIFACT_STORE",
+    "CAPABILITY_FILES_POSTBOX_REFERENCES",
+    "CAPABILITY_FILES_TABULAR_CONTENT",
+    "FileAccessProvider",
+    "ManagedArtifactRef",
+    "ManagedArtifactStore",
+    "ManagedArtifactWriteRequest",
+    "ManagedTabularFile",
+    "ManagedTabularFileAccessError",
+    "ManagedTabularFileContent",
+    "ManagedTabularFileError",
+    "ManagedTabularFileNotFoundError",
+    "ManagedTabularFileProvider",
+    "ManagedTabularFileUnavailableError",
+    "ManagedTabularFileValidationError",
+    "PostboxFileReferenceProvider",
+    "PostboxFileReferenceRef",
+    "PostboxFileReferenceRequest",
+    "managed_tabular_file_provider",
+    "postbox_file_reference_provider",
+]

src/govoplan_core/core/first_admin.py

@@ -356,7 +356,7 @@ def consume_first_admin_credential(
         tenant = Tenant(
             slug=clean_tenant_slug,
             name=clean_tenant_name,
-            default_locale="en",
+            default_locale="de",
             settings={},
             is_active=True,
         )

src/govoplan_core/core/form_evidence.py

@@ -0,0 +1,227 @@
+from __future__ import annotations
+
+from collections.abc import Mapping, Sequence
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Literal, Protocol, runtime_checkable
+
+from govoplan_core.core.institutional import EvidenceReference, InstitutionalReference
+
+
+CAPABILITY_FORM_EVIDENCE_PREFIX = "forms_runtime.evidence."
+
+FormEvidenceState = Literal[
+    "accepted",
+    "pending",
+    "rejected",
+    "expired",
+    "revoked",
+    "unavailable",
+]
+
+_FORM_EVIDENCE_STATES = {
+    "accepted",
+    "pending",
+    "rejected",
+    "expired",
+    "revoked",
+    "unavailable",
+}
+
+
+class FormEvidenceContractError(ValueError):
+    """Stable error for provider-neutral Form evidence operations."""
+
+
+@dataclass(frozen=True, slots=True)
+class FormEvidenceGrantRequest:
+    """Request a short-lived, purpose-bound grant from an evidence owner."""
+
+    tenant_id: str
+    instance_id: str
+    definition_ref: InstitutionalReference
+    evidence_kind: str
+    purpose: str
+    idempotency_key: str
+    expires_at: datetime
+    custodian_ref: str | None = None
+    max_size_bytes: int | None = None
+    allowed_content_types: tuple[str, ...] = ()
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text(self.tenant_id, "Form evidence tenant")
+        _require_text(self.instance_id, "Form evidence instance")
+        _require_text(self.evidence_kind, "Form evidence kind")
+        _require_text(self.purpose, "Form evidence purpose")
+        _require_text(self.idempotency_key, "Form evidence idempotency key")
+        if (
+            self.definition_ref.kind != "form"
+            or self.definition_ref.tenant_id != self.tenant_id
+            or not self.definition_ref.version
+        ):
+            raise FormEvidenceContractError(
+                "Form evidence grants require an exact same-tenant Form definition."
+            )
+        if self.expires_at.tzinfo is None or self.expires_at.utcoffset() is None:
+            raise FormEvidenceContractError(
+                "Form evidence grant expiry must include a timezone."
+            )
+        if self.max_size_bytes is not None and self.max_size_bytes <= 0:
+            raise FormEvidenceContractError(
+                "Form evidence grant size limits must be positive."
+            )
+        if any(not item.strip() for item in self.allowed_content_types):
+            raise FormEvidenceContractError(
+                "Form evidence content types cannot be empty."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class FormEvidenceGrant:
+    provider_id: str
+    grant_id: str
+    upload_token: str | None
+    upload_url: str
+    expires_at: datetime
+    max_size_bytes: int
+    allowed_content_types: tuple[str, ...] = ()
+    replayed: bool = False
+
+    def __post_init__(self) -> None:
+        for value, label in (
+            (self.provider_id, "Form evidence provider"),
+            (self.grant_id, "Form evidence grant"),
+            (self.upload_url, "Form evidence upload URL"),
+        ):
+            _require_text(value, label)
+        if self.upload_token is not None:
+            _require_text(self.upload_token, "Form evidence upload token")
+        if self.max_size_bytes <= 0:
+            raise FormEvidenceContractError(
+                "Form evidence grant size limits must be positive."
+            )
+        if self.expires_at.tzinfo is None or self.expires_at.utcoffset() is None:
+            raise FormEvidenceContractError(
+                "Form evidence grant expiry must include a timezone."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class FormEvidenceInspectionRequest:
+    tenant_id: str
+    instance_id: str
+    definition_ref: InstitutionalReference
+    evidence: EvidenceReference
+    purpose: str
+    final: bool
+
+    def __post_init__(self) -> None:
+        _require_text(self.tenant_id, "Form evidence tenant")
+        _require_text(self.instance_id, "Form evidence instance")
+        _require_text(self.purpose, "Form evidence purpose")
+        if self.definition_ref.tenant_id != self.tenant_id:
+            raise FormEvidenceContractError(
+                "Form evidence inspection cannot cross tenants."
+            )
+        if self.evidence.tenant_id != self.tenant_id:
+            raise FormEvidenceContractError(
+                "Form evidence inspection cannot cross tenants."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class FormEvidenceInspection:
+    provider_id: str
+    reference: EvidenceReference
+    state: FormEvidenceState
+    observed_at: datetime
+    retryable: bool = False
+    reason: str | None = None
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text(self.provider_id, "Form evidence provider")
+        if self.state not in _FORM_EVIDENCE_STATES:
+            raise FormEvidenceContractError(
+                f"Unsupported Form evidence state: {self.state!r}."
+            )
+        if self.observed_at.tzinfo is None or self.observed_at.utcoffset() is None:
+            raise FormEvidenceContractError(
+                "Form evidence inspection time must include a timezone."
+            )
+        if self.state == "accepted" and self.retryable:
+            raise FormEvidenceContractError(
+                "Accepted Form evidence cannot require a retry."
+            )
+
+    @property
+    def accepted(self) -> bool:
+        return self.state == "accepted"
+
+
+@runtime_checkable
+class FormEvidenceProvider(Protocol):
+    provider_id: str
+
+    def supported_kinds(self) -> Sequence[str]: ...
+
+    def create_upload_grant(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: FormEvidenceGrantRequest,
+    ) -> FormEvidenceGrant: ...
+
+    def inspect_evidence(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: FormEvidenceInspectionRequest,
+    ) -> FormEvidenceInspection: ...
+
+
+def form_evidence_capability(provider_id: str) -> str:
+    normalized = str(provider_id or "").strip().lower().replace("-", "_")
+    if not normalized or not normalized.replace("_", "").isalnum():
+        raise FormEvidenceContractError("Invalid Form evidence provider id.")
+    return f"{CAPABILITY_FORM_EVIDENCE_PREFIX}{normalized}"
+
+
+def form_evidence_provider(
+    registry: object | None,
+    provider_id: str,
+) -> FormEvidenceProvider | None:
+    capability_name = form_evidence_capability(provider_id)
+    if registry is None or not hasattr(registry, "has_capability"):
+        return None
+    if not registry.has_capability(capability_name):
+        return None
+    if hasattr(registry, "require_capability"):
+        provider = registry.require_capability(capability_name)
+    elif hasattr(registry, "capability"):
+        provider = registry.capability(capability_name)
+    else:
+        return None
+    return provider if isinstance(provider, FormEvidenceProvider) else None
+
+
+def _require_text(value: str, label: str) -> None:
+    if not isinstance(value, str) or not value.strip():
+        raise FormEvidenceContractError(f"{label} is required.")
+
+
+__all__ = [
+    "CAPABILITY_FORM_EVIDENCE_PREFIX",
+    "FormEvidenceContractError",
+    "FormEvidenceGrant",
+    "FormEvidenceGrantRequest",
+    "FormEvidenceInspection",
+    "FormEvidenceInspectionRequest",
+    "FormEvidenceProvider",
+    "FormEvidenceState",
+    "form_evidence_capability",
+    "form_evidence_provider",
+]

src/govoplan_core/core/information_governance.py

@@ -0,0 +1,236 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Literal, Mapping, cast
+
+
+InformationGovernanceAdoption = Literal[
+    "not_applicable",
+    "contract_only",
+    "partial",
+    "enforced",
+]
+
+INFORMATION_GOVERNANCE_ADOPTION_ORDER: tuple[InformationGovernanceAdoption, ...] = (
+    "not_applicable",
+    "contract_only",
+    "partial",
+    "enforced",
+)
+
+
+class InformationGovernanceDeclarationError(ValueError):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class InformationGovernanceDimension:
+    """Truthful module-level adoption claim for one cross-cutting dimension."""
+
+    adoption: InformationGovernanceAdoption = "contract_only"
+    object_types: tuple[str, ...] = ()
+    evidence: tuple[str, ...] = ()
+    limitation: str | None = None
+
+    def __post_init__(self) -> None:
+        if self.adoption not in INFORMATION_GOVERNANCE_ADOPTION_ORDER:
+            raise InformationGovernanceDeclarationError(
+                f"Unsupported information-governance adoption: {self.adoption!r}."
+            )
+        for field_name in ("object_types", "evidence"):
+            values = getattr(self, field_name)
+            if len(values) != len(set(values)) or any(not item.strip() for item in values):
+                raise InformationGovernanceDeclarationError(
+                    f"Information-governance {field_name.replace('_', ' ')} must "
+                    "contain unique non-empty values."
+                )
+        if self.adoption == "enforced" and not self.evidence:
+            raise InformationGovernanceDeclarationError(
+                "An enforced information-governance dimension requires evidence."
+            )
+        if self.adoption in {"partial", "enforced"} and not self.object_types:
+            raise InformationGovernanceDeclarationError(
+                "Partial and enforced information-governance dimensions must "
+                "name their covered object types."
+            )
+        if self.adoption == "not_applicable" and self.object_types:
+            raise InformationGovernanceDeclarationError(
+                "A non-applicable information-governance dimension cannot declare object types."
+            )
+        if self.adoption in {"contract_only", "partial"} and not str(
+            self.limitation or ""
+        ).strip():
+            raise InformationGovernanceDeclarationError(
+                "Contract-only and partial adoption must state the current limitation."
+            )
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "adoption": self.adoption,
+            "object_types": list(self.object_types),
+            "evidence": list(self.evidence),
+            "limitation": self.limitation,
+        }
+
+
+def _contract_only_dimension() -> InformationGovernanceDimension:
+    return InformationGovernanceDimension(
+        adoption="contract_only",
+        limitation=(
+            "The platform contract applies, but module-specific adoption evidence "
+            "has not been declared."
+        ),
+    )
+
+
+@dataclass(frozen=True, slots=True)
+class ModuleInformationGovernance:
+    """Cross-cutting data-use requirements and honest adoption evidence."""
+
+    temporal_browsing: InformationGovernanceDimension = field(
+        default_factory=_contract_only_dimension
+    )
+    purpose_aware_access: InformationGovernanceDimension = field(
+        default_factory=_contract_only_dimension
+    )
+    retention: InformationGovernanceDimension = field(
+        default_factory=_contract_only_dimension
+    )
+    institutional_context: InformationGovernanceDimension = field(
+        default_factory=_contract_only_dimension
+    )
+    current_authorization_for_historical_reads: bool = True
+    contract_version: str = "1"
+
+    def __post_init__(self) -> None:
+        if self.contract_version != "1":
+            raise InformationGovernanceDeclarationError(
+                "Unsupported module information-governance contract version."
+            )
+        if not self.current_authorization_for_historical_reads:
+            raise InformationGovernanceDeclarationError(
+                "Historical reads must always use current authorization."
+            )
+        for name, dimension in self.dimensions.items():
+            if not isinstance(dimension, InformationGovernanceDimension):
+                raise InformationGovernanceDeclarationError(
+                    f"Information-governance dimension {name!r} has an invalid value."
+                )
+
+    @property
+    def dimensions(self) -> Mapping[str, InformationGovernanceDimension]:
+        return {
+            "temporal_browsing": self.temporal_browsing,
+            "purpose_aware_access": self.purpose_aware_access,
+            "retention": self.retention,
+            "institutional_context": self.institutional_context,
+        }
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "contract_version": self.contract_version,
+            "current_authorization_for_historical_reads": (
+                self.current_authorization_for_historical_reads
+            ),
+            "dimensions": {
+                name: dimension.to_dict()
+                for name, dimension in self.dimensions.items()
+            },
+        }
+
+
+def information_governance_from_mapping(
+    value: Mapping[str, object],
+) -> ModuleInformationGovernance:
+    raw_dimensions = value.get("dimensions")
+    if not isinstance(raw_dimensions, Mapping):
+        raise InformationGovernanceDeclarationError(
+            "Information-governance dimensions must be an object."
+        )
+
+    def dimension(name: str) -> InformationGovernanceDimension:
+        raw_dimension = raw_dimensions.get(name)
+        if not isinstance(raw_dimension, Mapping):
+            raise InformationGovernanceDeclarationError(
+                f"Information-governance dimension {name!r} must be an object."
+            )
+
+        def text_tuple(field_name: str) -> tuple[str, ...]:
+            raw_values = raw_dimension.get(field_name, ())
+            if not isinstance(raw_values, (list, tuple)):
+                raise InformationGovernanceDeclarationError(
+                    f"Information-governance {name}.{field_name} must be a list."
+                )
+            if any(not isinstance(item, str) for item in raw_values):
+                raise InformationGovernanceDeclarationError(
+                    f"Information-governance {name}.{field_name} must contain strings."
+                )
+            return tuple(raw_values)
+
+        raw_limitation = raw_dimension.get("limitation")
+        raw_adoption = raw_dimension.get("adoption") or "contract_only"
+        if not isinstance(raw_adoption, str):
+            raise InformationGovernanceDeclarationError(
+                f"Information-governance {name}.adoption must be a string."
+            )
+        if raw_limitation is not None and not isinstance(raw_limitation, str):
+            raise InformationGovernanceDeclarationError(
+                f"Information-governance {name}.limitation must be a string."
+            )
+        return InformationGovernanceDimension(
+            adoption=cast(
+                InformationGovernanceAdoption,
+                raw_adoption,
+            ),
+            object_types=text_tuple("object_types"),
+            evidence=text_tuple("evidence"),
+            limitation=raw_limitation,
+        )
+
+    current_authorization = value.get(
+        "current_authorization_for_historical_reads",
+        True,
+    )
+    if not isinstance(current_authorization, bool):
+        raise InformationGovernanceDeclarationError(
+            "current_authorization_for_historical_reads must be boolean."
+        )
+    return ModuleInformationGovernance(
+        contract_version=str(value.get("contract_version") or "1"),
+        current_authorization_for_historical_reads=current_authorization,
+        temporal_browsing=dimension("temporal_browsing"),
+        purpose_aware_access=dimension("purpose_aware_access"),
+        retention=dimension("retention"),
+        institutional_context=dimension("institutional_context"),
+    )
+
+
+def information_governance_maturity_issues(
+    declaration: ModuleInformationGovernance,
+    *,
+    maturity: str | None,
+) -> tuple[str, ...]:
+    if maturity not in {"reference_ready", "supported", "lts"}:
+        return ()
+    incomplete = [
+        name
+        for name, dimension in declaration.dimensions.items()
+        if dimension.adoption not in {"not_applicable", "enforced"}
+    ]
+    if not incomplete:
+        return ()
+    return (
+        f"Maturity {maturity!r} requires enforced or explicitly non-applicable "
+        "information governance for: " + ", ".join(incomplete),
+    )
+
+
+__all__ = [
+    "INFORMATION_GOVERNANCE_ADOPTION_ORDER",
+    "InformationGovernanceAdoption",
+    "InformationGovernanceDeclarationError",
+    "InformationGovernanceDimension",
+    "ModuleInformationGovernance",
+    "information_governance_from_mapping",
+    "information_governance_maturity_issues",
+]

src/govoplan_core/core/infrastructure_capabilities.py

@@ -0,0 +1,614 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass
+from datetime import UTC, datetime
+import json
+import os
+from pathlib import Path
+import re
+from typing import Any, Protocol, runtime_checkable
+
+
+DEPLOYMENT_CAPABILITIES_ENV = "GOVOPLAN_DEPLOYMENT_CAPABILITIES_PATH"
+INFRASTRUCTURE_DEPENDENCY_PROVIDER_CAPABILITY_PREFIX = (
+    "infrastructure.dependency_inventory."
+)
+MAX_CAPABILITY_DOCUMENT_BYTES = 256 * 1024
+CAPABILITY_STATES = frozenset(
+    {
+        "configured",
+        "available_unconfigured",
+        "externally_supplied",
+        "unavailable",
+    }
+)
+_ENV_REFERENCE_RE = re.compile(r"^env:[A-Za-z_][A-Za-z0-9_]*$")
+_DEPENDENCY_STATES = frozenset(
+    {"active", "inactive", "data_present", "pending_work", "runtime_binding"}
+)
+
+
+class InfrastructureCapabilityReceiptError(ValueError):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructureDependency:
+    """A non-secret module-owned dependency on deployment infrastructure."""
+
+    capability_id: str
+    module_id: str
+    dependency_type: str
+    dependency_ref: str
+    state: str
+    scope: str
+    summary: str
+    metrics: Mapping[str, int]
+    required_action: str
+
+    def __post_init__(self) -> None:
+        for field_name, value, maximum in (
+            ("capability_id", self.capability_id, 120),
+            ("module_id", self.module_id, 120),
+            ("dependency_type", self.dependency_type, 120),
+            ("dependency_ref", self.dependency_ref, 240),
+            ("scope", self.scope, 120),
+            ("summary", self.summary, 1000),
+            ("required_action", self.required_action, 1000),
+        ):
+            if (
+                not value.strip()
+                or len(value) > maximum
+                or any(ord(char) < 32 for char in value)
+            ):
+                raise ValueError(
+                    f"Infrastructure dependency {field_name} is invalid."
+                )
+        if self.state not in _DEPENDENCY_STATES:
+            raise ValueError(
+                f"Infrastructure dependency state is unsupported: {self.state!r}."
+            )
+        if len(self.metrics) > 20 or any(
+            not isinstance(key, str)
+            or not key.strip()
+            or len(key) > 80
+            or any(ord(char) < 32 for char in key)
+            or type(value) is not int
+            or value < 0
+            for key, value in self.metrics.items()
+        ):
+            raise ValueError("Infrastructure dependency metrics are invalid.")
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "capability_id": self.capability_id,
+            "module_id": self.module_id,
+            "dependency_type": self.dependency_type,
+            "dependency_ref": self.dependency_ref,
+            "state": self.state,
+            "scope": self.scope,
+            "summary": self.summary,
+            "metrics": dict(sorted(self.metrics.items())),
+            "required_action": self.required_action,
+        }
+
+
+@runtime_checkable
+class InfrastructureDependencyProvider(Protocol):
+    module_id: str
+    capability_ids: tuple[str, ...]
+
+    def infrastructure_dependencies(self) -> tuple[InfrastructureDependency, ...]:
+        ...
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructureDependencyProviderReport:
+    module_id: str
+    capability_ids: tuple[str, ...]
+    state: str
+    dependency_count: int
+    error: str | None = None
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "module_id": self.module_id,
+            "capability_ids": list(self.capability_ids),
+            "state": self.state,
+            "dependency_count": self.dependency_count,
+            "error": self.error,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructureDependencyInventory:
+    installation_id: str
+    generated_at: str
+    complete: bool
+    inspected_capability_ids: tuple[str, ...]
+    providers: tuple[InfrastructureDependencyProviderReport, ...]
+    dependencies: tuple[InfrastructureDependency, ...]
+    schema_version: int = 1
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "schema_version": self.schema_version,
+            "installation_id": self.installation_id,
+            "generated_at": self.generated_at,
+            "complete": self.complete,
+            "inspected_capability_ids": list(self.inspected_capability_ids),
+            "providers": [item.to_dict() for item in self.providers],
+            "dependencies": [item.to_dict() for item in self.dependencies],
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructureCapability:
+    id: str
+    label: str
+    state: str
+    source: str
+    detail: str
+    endpoint: Mapping[str, object]
+    secret_refs: tuple[str, ...]
+    dependent_modules: tuple[str, ...]
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "id": self.id,
+            "label": self.label,
+            "state": self.state,
+            "source": self.source,
+            "detail": self.detail,
+            "endpoint": dict(self.endpoint),
+            "secret_refs": list(self.secret_refs),
+            "dependent_modules": list(self.dependent_modules),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructurePostInstallTask:
+    id: str
+    resume_key: str
+    capability_id: str
+    state: str
+    owner_module: str
+    summary: str
+    required_inputs: tuple[str, ...]
+    secret_boundary: str
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "id": self.id,
+            "resume_key": self.resume_key,
+            "capability_id": self.capability_id,
+            "state": self.state,
+            "owner_module": self.owner_module,
+            "summary": self.summary,
+            "required_inputs": list(self.required_inputs),
+            "secret_boundary": self.secret_boundary,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class InfrastructureCapabilityReceipt:
+    installation_id: str
+    profile: str
+    capabilities: tuple[InfrastructureCapability, ...]
+    post_install_tasks: tuple[InfrastructurePostInstallTask, ...]
+    schema_version: int = 1
+
+    def capability(self, capability_id: str) -> InfrastructureCapability | None:
+        return next(
+            (item for item in self.capabilities if item.id == capability_id),
+            None,
+        )
+
+    def tasks_for(
+        self,
+        *,
+        capability_id: str | None = None,
+        owner_module: str | None = None,
+    ) -> tuple[InfrastructurePostInstallTask, ...]:
+        return tuple(
+            item
+            for item in self.post_install_tasks
+            if (capability_id is None or item.capability_id == capability_id)
+            and (owner_module is None or item.owner_module == owner_module)
+        )
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "schema_version": self.schema_version,
+            "installation_id": self.installation_id,
+            "profile": self.profile,
+            "capabilities": [item.to_dict() for item in self.capabilities],
+            "post_install_tasks": [
+                item.to_dict() for item in self.post_install_tasks
+            ],
+        }
+
+
+def load_infrastructure_capability_receipt(
+    path: Path | str | None = None,
+) -> InfrastructureCapabilityReceipt | None:
+    configured_path = path
+    if configured_path is None:
+        raw_path = os.getenv(DEPLOYMENT_CAPABILITIES_ENV, "").strip()
+        if not raw_path:
+            return None
+        configured_path = raw_path
+    return read_infrastructure_capability_receipt(Path(configured_path))
+
+
+def read_infrastructure_capability_receipt(
+    path: Path,
+) -> InfrastructureCapabilityReceipt:
+    if path.is_symlink() or not path.is_file():
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt is not a regular file."
+        )
+    try:
+        expected_size = path.stat().st_size
+        if expected_size > MAX_CAPABILITY_DOCUMENT_BYTES:
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability receipt exceeds 256 KiB."
+            )
+        raw = path.read_bytes()
+    except OSError as exc:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt could not be read."
+        ) from exc
+    if len(raw) != expected_size:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt changed while being read."
+        )
+    try:
+        payload = json.loads(raw.decode("utf-8"))
+    except (UnicodeDecodeError, json.JSONDecodeError) as exc:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt is not valid UTF-8 JSON."
+        ) from exc
+    return infrastructure_capability_receipt_from_mapping(payload)
+
+
+def infrastructure_capability_receipt_from_mapping(
+    payload: object,
+) -> InfrastructureCapabilityReceipt:
+    if (
+        not isinstance(payload, Mapping)
+        or type(payload.get("schema_version")) is not int
+        or payload.get("schema_version") != 1
+    ):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt has an unsupported schema."
+        )
+    raw_capabilities = payload.get("capabilities")
+    if not isinstance(raw_capabilities, list) or len(raw_capabilities) > 100:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt has invalid capabilities."
+        )
+    capabilities = tuple(_capability(item) for item in raw_capabilities)
+    capability_ids = [item.id for item in capabilities]
+    if len(capability_ids) != len(set(capability_ids)):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt repeats a capability id."
+        )
+    raw_tasks = payload.get("post_install_tasks", [])
+    if not isinstance(raw_tasks, list) or len(raw_tasks) > 100:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability receipt has invalid post-install tasks."
+        )
+    tasks = tuple(_task(item) for item in raw_tasks)
+    known_capability_ids = set(capability_ids)
+    if any(item.capability_id not in known_capability_ids for item in tasks):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment post-install task references an unknown capability."
+        )
+    return InfrastructureCapabilityReceipt(
+        installation_id=_required_text(payload, "installation_id", maximum=100),
+        profile=_required_text(payload, "profile", maximum=100),
+        capabilities=capabilities,
+        post_install_tasks=tasks,
+    )
+
+
+def deployment_capability_status(
+    path: Path | str | None = None,
+) -> dict[str, object]:
+    try:
+        receipt = load_infrastructure_capability_receipt(path)
+    except InfrastructureCapabilityReceiptError as exc:
+        return _unavailable_status(configured=True, error=str(exc))
+    if receipt is None:
+        return _unavailable_status(configured=False, error=None)
+    return {
+        "configured": True,
+        "available": True,
+        **receipt.to_dict(),
+        "error": None,
+    }
+
+
+def collect_infrastructure_dependency_inventory(
+    registry: object,
+    *,
+    installation_id: str,
+    observed_at: datetime | None = None,
+) -> InfrastructureDependencyInventory:
+    """Collect actual module-owned dependencies without importing module internals."""
+
+    normalized_installation_id = installation_id.strip()
+    if not normalized_installation_id or len(normalized_installation_id) > 100:
+        raise ValueError("Infrastructure dependency installation id is invalid.")
+    capability_names = getattr(registry, "capability_names", None)
+    capability = getattr(registry, "capability", None)
+    if not callable(capability_names) or not callable(capability):
+        raise ValueError("Infrastructure dependency inventory requires a module registry.")
+
+    provider_names = tuple(
+        name
+        for name in capability_names()
+        if isinstance(name, str)
+        and name.startswith(INFRASTRUCTURE_DEPENDENCY_PROVIDER_CAPABILITY_PREFIX)
+    )
+    reports: list[InfrastructureDependencyProviderReport] = []
+    dependencies: list[InfrastructureDependency] = []
+    inspected_capability_ids: set[str] = set()
+    complete = True
+
+    for provider_name in sorted(provider_names):
+        expected_module_id = provider_name.removeprefix(
+            INFRASTRUCTURE_DEPENDENCY_PROVIDER_CAPABILITY_PREFIX
+        )
+        module_id = expected_module_id or "unknown"
+        declared_ids: tuple[str, ...] = ()
+        try:
+            provider = capability(provider_name)
+            if not isinstance(provider, InfrastructureDependencyProvider):
+                raise TypeError("provider does not implement the inventory contract")
+            module_id = provider.module_id.strip()
+            declared_ids = tuple(
+                sorted(
+                    {
+                        item.strip()
+                        for item in provider.capability_ids
+                        if isinstance(item, str) and item.strip()
+                    }
+                )
+            )
+            if (
+                module_id != expected_module_id
+                or len(module_id) > 120
+                or any(ord(char) < 32 for char in module_id)
+                or not declared_ids
+                or len(declared_ids) > 30
+                or any(
+                    len(item) > 120 or any(ord(char) < 32 for char in item)
+                    for item in declared_ids
+                )
+            ):
+                raise ValueError("provider identity or capability declaration is invalid")
+            provider_dependencies = tuple(provider.infrastructure_dependencies())
+            if len(provider_dependencies) > 10_000:
+                raise ValueError("provider dependency inventory is too large")
+            seen_refs: set[tuple[str, str, str]] = set()
+            for item in provider_dependencies:
+                if not isinstance(item, InfrastructureDependency):
+                    raise TypeError("provider returned an invalid dependency")
+                if item.module_id != module_id or item.capability_id not in declared_ids:
+                    raise ValueError("provider returned a dependency outside its declaration")
+                identity = (
+                    item.capability_id,
+                    item.dependency_type,
+                    item.dependency_ref,
+                )
+                if identity in seen_refs:
+                    raise ValueError("provider returned a duplicate dependency")
+                seen_refs.add(identity)
+            if len(dependencies) + len(provider_dependencies) > 10_000:
+                raise ValueError("combined dependency inventory is too large")
+            dependencies.extend(provider_dependencies)
+            inspected_capability_ids.update(declared_ids)
+            reports.append(
+                InfrastructureDependencyProviderReport(
+                    module_id=module_id,
+                    capability_ids=declared_ids,
+                    state="complete",
+                    dependency_count=len(provider_dependencies),
+                )
+            )
+        except Exception as exc:
+            complete = False
+            inspected_capability_ids.update(declared_ids)
+            reports.append(
+                InfrastructureDependencyProviderReport(
+                    module_id=module_id,
+                    capability_ids=declared_ids,
+                    state="error",
+                    dependency_count=0,
+                    error=(
+                        f"{type(exc).__name__}: provider inventory could not be completed"
+                    ),
+                )
+            )
+
+    timestamp = observed_at or datetime.now(UTC)
+    if timestamp.tzinfo is None:
+        timestamp = timestamp.replace(tzinfo=UTC)
+    return InfrastructureDependencyInventory(
+        installation_id=normalized_installation_id,
+        generated_at=timestamp.astimezone(UTC).isoformat(),
+        complete=complete,
+        inspected_capability_ids=tuple(sorted(inspected_capability_ids)),
+        providers=tuple(
+            sorted(reports, key=lambda item: (item.module_id, item.capability_ids))
+        ),
+        dependencies=tuple(
+            sorted(
+                dependencies,
+                key=lambda item: (
+                    item.capability_id,
+                    item.module_id,
+                    item.dependency_type,
+                    item.dependency_ref,
+                ),
+            )
+        ),
+    )
+
+
+def _capability(value: object) -> InfrastructureCapability:
+    if not isinstance(value, Mapping):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability entries must be objects."
+        )
+    state = _required_text(value, "state", maximum=40)
+    if state not in CAPABILITY_STATES:
+        raise InfrastructureCapabilityReceiptError(
+            f"Deployment capability state is unsupported: {state!r}."
+        )
+    normalized_endpoint = _normalized_endpoint(value.get("endpoint", {}))
+    secret_refs = _string_list(value.get("secret_refs"), maximum_items=30)
+    if any(not _ENV_REFERENCE_RE.fullmatch(item) for item in secret_refs):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability secrets must use environment references."
+        )
+    return InfrastructureCapability(
+        id=_required_text(value, "id", maximum=120),
+        label=_required_text(value, "label", maximum=200),
+        state=state,
+        source=_required_text(value, "source", maximum=120),
+        detail=_required_text(value, "detail", maximum=1000),
+        endpoint=normalized_endpoint,
+        secret_refs=secret_refs,
+        dependent_modules=_string_list(
+            value.get("dependent_modules"),
+            maximum_items=100,
+        ),
+    )
+
+
+def _normalized_endpoint(value: object) -> dict[str, object]:
+    if not isinstance(value, Mapping) or len(value) > 10:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability endpoint metadata is invalid."
+        )
+    endpoint: dict[str, object] = {}
+    for key, raw in value.items():
+        if not isinstance(key, str) or not key or len(key) > 50:
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability endpoint key is invalid."
+            )
+        if any(
+            marker in key.casefold()
+            for marker in ("password", "secret", "token", "credential")
+        ):
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability endpoint metadata contains a secret field."
+            )
+        if key.casefold() == "port" and (
+            type(raw) is not int or not 1 <= raw <= 65535
+        ):
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability endpoint port is invalid."
+            )
+        if isinstance(raw, bool) or raw is None:
+            endpoint[key] = raw
+        elif isinstance(raw, int):
+            endpoint[key] = raw
+        elif isinstance(raw, str) and len(raw) <= 500:
+            endpoint[key] = raw
+        else:
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability endpoint value is invalid."
+            )
+    return endpoint
+
+
+def _task(value: object) -> InfrastructurePostInstallTask:
+    if not isinstance(value, Mapping):
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment post-install task entries must be objects."
+        )
+    return InfrastructurePostInstallTask(
+        id=_required_text(value, "id", maximum=120),
+        resume_key=_required_text(value, "resume_key", maximum=240),
+        capability_id=_required_text(value, "capability_id", maximum=120),
+        state=_required_text(value, "state", maximum=40),
+        owner_module=_required_text(value, "owner_module", maximum=120),
+        summary=_required_text(value, "summary", maximum=1000),
+        required_inputs=_string_list(
+            value.get("required_inputs"),
+            maximum_items=30,
+        ),
+        secret_boundary=_required_text(
+            value,
+            "secret_boundary",
+            maximum=120,
+        ),
+    )
+
+
+def _required_text(
+    value: Mapping[str, Any],
+    key: str,
+    *,
+    maximum: int,
+) -> str:
+    raw = value.get(key)
+    text = str(raw).strip() if raw is not None else ""
+    if not text or len(text) > maximum:
+        raise InfrastructureCapabilityReceiptError(
+            f"Deployment capability field {key!r} is invalid."
+        )
+    return text
+
+
+def _string_list(value: object, *, maximum_items: int) -> tuple[str, ...]:
+    if not isinstance(value, list) or len(value) > maximum_items:
+        raise InfrastructureCapabilityReceiptError(
+            "Deployment capability list field is invalid."
+        )
+    result: list[str] = []
+    for item in value:
+        if not isinstance(item, str) or not item.strip() or len(item) > 500:
+            raise InfrastructureCapabilityReceiptError(
+                "Deployment capability list item is invalid."
+            )
+        result.append(item.strip())
+    return tuple(result)
+
+
+def _unavailable_status(*, configured: bool, error: str | None) -> dict[str, object]:
+    return {
+        "configured": configured,
+        "available": False,
+        "schema_version": None,
+        "installation_id": None,
+        "profile": None,
+        "capabilities": [],
+        "post_install_tasks": [],
+        "error": error,
+    }
+
+
+__all__ = [
+    "CAPABILITY_STATES",
+    "DEPLOYMENT_CAPABILITIES_ENV",
+    "INFRASTRUCTURE_DEPENDENCY_PROVIDER_CAPABILITY_PREFIX",
+    "InfrastructureCapability",
+    "InfrastructureCapabilityReceipt",
+    "InfrastructureCapabilityReceiptError",
+    "InfrastructureDependency",
+    "InfrastructureDependencyInventory",
+    "InfrastructureDependencyProvider",
+    "InfrastructureDependencyProviderReport",
+    "InfrastructurePostInstallTask",
+    "collect_infrastructure_dependency_inventory",
+    "deployment_capability_status",
+    "infrastructure_capability_receipt_from_mapping",
+    "load_infrastructure_capability_receipt",
+    "read_infrastructure_capability_receipt",
+]

src/govoplan_core/core/install_config.py

@@ -698,12 +698,15 @@ def _validate_module_catalog_trust(
             "A module catalog source is configured without a trusted keyring file.",
             "Pin the published GovOPlaN catalog keyring locally and set GOVOPLAN_MODULE_PACKAGE_CATALOG_TRUSTED_KEYS_FILE.",
         )
-    if not _clean(env.get("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL")):
+    if not (
+        _clean(env.get("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS"))
+        or _clean(env.get("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL"))
+    ):
         collector.add(
             "error",
-            "GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL",
+            "GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS",
             "A module catalog source is configured without an approved release channel.",
-            "Set GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL=stable or another approved deployment channel.",
+            "Set GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS=stable or another approved deployment channel.",
         )
 
 
@@ -734,6 +737,8 @@ CALENDAR_OUTBOX_TERMINAL_RETENTION_DAYS=90
 PLATFORM_EVENT_OUTBOX_MAX_ATTEMPTS=8
 PLATFORM_EVENT_OUTBOX_TERMINAL_RETENTION_DAYS=90
 SCHEDULING_CANCELLATION_NOTICE_DAYS=30
+SCHEDULING_PUBLIC_SELF_ENROLLMENT_ENABLED=true
+SCHEDULING_PUBLIC_SELF_ENROLLMENT_MAX_CAPACITY=10000
 
 # Deployment-wide connector egress policy. Enable private networks only when
 # this installation intentionally integrates with internal services.
@@ -777,7 +782,7 @@ DEV_MAILBOX_API_ENABLED=false
 
 GOVOPLAN_MODULE_PACKAGE_CATALOG_URL=https://govoplan.add-ideas.de/catalogs/v1/channels/stable.json
 GOVOPLAN_MODULE_PACKAGE_CATALOG_TRUSTED_KEYS_FILE=/etc/govoplan/catalog-keyring.json
-GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL=stable
+GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS=stable
 """
 
 
@@ -815,6 +820,8 @@ CALENDAR_OUTBOX_TERMINAL_RETENTION_DAYS=90
 PLATFORM_EVENT_OUTBOX_MAX_ATTEMPTS=8
 PLATFORM_EVENT_OUTBOX_TERMINAL_RETENTION_DAYS=90
 SCHEDULING_CANCELLATION_NOTICE_DAYS=30
+SCHEDULING_PUBLIC_SELF_ENROLLMENT_ENABLED=true
+SCHEDULING_PUBLIC_SELF_ENROLLMENT_MAX_CAPACITY=10000
 
 GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS=true
 GOVOPLAN_CONNECTOR_MAX_STRUCTURED_RESPONSE_BYTES=16777216

src/govoplan_core/core/mail.py

@@ -9,6 +9,7 @@ from typing import Protocol, runtime_checkable
 CAPABILITY_MAIL_DELIVERY_OUTBOX = "mail.delivery_outbox"
 CAPABILITY_MAIL_NOTIFICATION_DELIVERY = "mail.notificationDelivery"
 CAPABILITY_MAIL_BOUNCE_PROCESSING = "mail.bounce_processing"
+CAPABILITY_MAIL_POSTBOX_BRIDGE = "mail.postbox_bridge"
 
 
 @dataclass(frozen=True, slots=True)
@@ -81,6 +82,28 @@ class MailBounceObservationRef:
     evidence: Mapping[str, object] = field(default_factory=dict)
 
 
+@dataclass(frozen=True, slots=True)
+class MailPostboxBridgeRequest:
+    tenant_id: str
+    target: object
+    profile_id: str
+    folder: str
+    uid: str
+    uidvalidity: str
+    raw_message: bytes
+    classification: str = "internal"
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+
+@dataclass(frozen=True, slots=True)
+class MailPostboxBridgeResult:
+    postbox_id: str
+    message_id: str
+    delivery_id: str
+    duplicate: bool
+    source_digest: str
+
+
 @runtime_checkable
 class MailBounceProcessingProvider(Protocol):
     """Mail-owned DSN ingestion and durable correlation boundary."""
@@ -116,6 +139,17 @@ class MailBounceProcessingProvider(Protocol):
         ...
 
 
+@runtime_checkable
+class MailPostboxBridgeProvider(Protocol):
+    """Translate one immutable Mail observation into Postbox delivery."""
+
+    def bridge_message(
+        self,
+        session: object,
+        request: MailPostboxBridgeRequest,
+    ) -> MailPostboxBridgeResult: ...
+
+
 def notification_mail_delivery_provider(
     registry: object | None,
 ) -> NotificationMailDeliveryProvider | None:
@@ -150,3 +184,21 @@ def mail_bounce_processing_provider(
             "MailBounceProcessingProvider"
         )
     return provider
+
+
+def mail_postbox_bridge_provider(
+    registry: object | None,
+) -> MailPostboxBridgeProvider | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(CAPABILITY_MAIL_POSTBOX_BRIDGE)
+    ):
+        return None
+    provider = registry.require_capability(CAPABILITY_MAIL_POSTBOX_BRIDGE)
+    if not isinstance(provider, MailPostboxBridgeProvider):
+        raise TypeError(
+            "mail.postbox_bridge provider does not implement "
+            "MailPostboxBridgeProvider"
+        )
+    return provider

src/govoplan_core/core/module_installer.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from collections import defaultdict
 from collections.abc import Iterable, Mapping
 from contextlib import AbstractContextManager, closing
-from dataclasses import dataclass, field
+from dataclasses import dataclass, field, replace
 from datetime import UTC, datetime
 from importlib import metadata
 import hashlib
@@ -18,6 +18,7 @@ import sqlite3
 import stat
 import subprocess  # nosec B404 - installer commands are structured and policy-validated before execution.
 import sys
+import tempfile
 import tomllib
 from typing import Any, Literal
 import time
@@ -62,6 +63,7 @@ MIGRATION_TASK_PHASES = (
 MIGRATION_TASK_MUTATING_PHASES = {"pre_migration_prepare", "post_migration_backfill"}
 MIGRATION_TASK_REVIEW_SAFETY = {"requires_review", "forward_only", "destructive"}
 MIGRATION_TASK_BLOCKING_SAFETY = {"forward_only", "destructive"}
+MAX_PACKAGE_ARTIFACT_BYTES = 512 * 1024 * 1024
 
 
 @dataclass(frozen=True, slots=True)
@@ -464,14 +466,20 @@ def _package_target_action_preflight_issues(
             "Python installs must include the distribution package name so rollback can uninstall newly added packages.",
             item.module_id,
         ))
-    if item.python_ref and not _looks_pinned_dependency_ref(item.python_ref):
+    if item.python_ref and not (
+        _looks_pinned_dependency_ref(item.python_ref)
+        or _artifact_ref_is_digest_pinned(item, "python", item.python_ref)
+    ):
         issues.append(ModuleInstallerIssue(
             "blocker",
             "unpinned_python_ref",
             "Python install refs must be pinned to an exact version or tagged git ref.",
             item.module_id,
         ))
-    if item.webui_ref and not _looks_pinned_dependency_ref(item.webui_ref):
+    if item.webui_ref and not (
+        _looks_pinned_dependency_ref(item.webui_ref)
+        or _artifact_ref_is_digest_pinned(item, "webui", item.webui_ref)
+    ):
         issues.append(ModuleInstallerIssue(
             "blocker",
             "unpinned_webui_ref",
@@ -481,6 +489,23 @@ def _package_target_action_preflight_issues(
     return tuple(issues)
 
 
+def _artifact_ref_is_digest_pinned(
+    item: ModuleInstallPlanItem,
+    kind: str,
+    package_ref: str,
+) -> bool:
+    metadata = _artifact_metadata(item.artifact_integrity, kind)
+    if metadata is None:
+        return False
+    expected_ref = _artifact_text(metadata, "ref") or _artifact_text(metadata, "expected_ref")
+    sha256 = _artifact_text(metadata, "sha256")
+    return bool(
+        expected_ref == package_ref
+        and sha256
+        and re.fullmatch(r"[0-9a-f]{64}", sha256.lower())
+    )
+
+
 def _frontend_rebuild_preflight_issues(
     *,
     frontend_rebuild_required: bool,
@@ -519,6 +544,7 @@ def run_module_install_plan(
 ) -> ModuleInstallerRunResult:
     maintenance_mode = saved_maintenance_mode(session)
     effective_runtime_dir = runtime_dir or default_installer_runtime_dir(database_url)
+    effective_plan = plan
     preflight = module_install_preflight(
         plan=plan,
         available=available,
@@ -532,9 +558,30 @@ def run_module_install_plan(
     if not preflight.allowed:
         raise ModuleInstallerError("Install preflight is blocked: " + "; ".join(issue.message for issue in preflight.issues if issue.severity == "blocker"))
 
+    if not dry_run:
+        effective_plan = acquire_catalog_package_artifacts(
+            plan,
+            runtime_dir=effective_runtime_dir,
+        )
+        preflight = module_install_preflight(
+            plan=effective_plan,
+            available=available,
+            current_enabled=current_enabled,
+            desired_enabled=desired_enabled,
+            maintenance_mode=maintenance_mode.enabled,
+            session=session,
+            webui_root=webui_root,
+            runtime_dir=effective_runtime_dir,
+        )
+        if not preflight.allowed:
+            raise ModuleInstallerError(
+                "Install preflight is blocked after artifact acquisition: "
+                + "; ".join(issue.message for issue in preflight.issues if issue.severity == "blocker")
+            )
+
     state = _prepare_module_install_run(
         session=session,
-        plan=plan,
+        plan=effective_plan,
         preflight=preflight,
         database_url=database_url,
         effective_runtime_dir=effective_runtime_dir,
@@ -556,7 +603,7 @@ def run_module_install_plan(
 
     executed, failed_error = _execute_module_install_run(
         session=session,
-        plan=plan,
+        plan=effective_plan,
         available=available,
         effective_runtime_dir=effective_runtime_dir,
         state=state,
@@ -566,7 +613,7 @@ def run_module_install_plan(
         return _failed_module_install_run_result(
             session=session,
             state=state,
-            plan=plan,
+            plan=effective_plan,
             executed=executed,
             failed_error=failed_error,
             effective_runtime_dir=effective_runtime_dir,
@@ -578,7 +625,7 @@ def run_module_install_plan(
 
     return _applied_module_install_run_result(
         session=session,
-        plan=plan,
+        plan=effective_plan,
         desired_enabled=desired_enabled,
         activate_installed_modules=activate_installed_modules,
         remove_uninstalled_modules_from_desired=remove_uninstalled_modules_from_desired,
@@ -1568,9 +1615,11 @@ def _structured_item_commands(
     webui_changed = False
     if item.action in PACKAGE_TARGET_ACTIONS:
         if item.python_ref:
-            commands.append(_structured_command([sys.executable, "-m", "pip", "install", item.python_ref], source="module-plan.python"))
+            python_source = _verified_artifact_install_ref(item, "python") or item.python_ref
+            commands.append(_structured_command([sys.executable, "-m", "pip", "install", python_source], source="module-plan.python"))
         if item.webui_package and item.webui_ref and webui_root is not None:
-            commands.append(_structured_command([npm_bin, "pkg", "set", f"dependencies.{item.webui_package}={item.webui_ref}"], cwd=webui_root, source="module-plan.webui"))
+            webui_source = _verified_artifact_install_ref(item, "webui") or item.webui_ref
+            commands.append(_structured_command([npm_bin, "pkg", "set", f"dependencies.{item.webui_package}={webui_source}"], cwd=webui_root, source="module-plan.webui"))
             webui_changed = True
     elif item.action == "uninstall":
         if item.python_package:
@@ -1581,6 +1630,14 @@ def _structured_item_commands(
     return tuple(commands), webui_changed
 
 
+def _verified_artifact_install_ref(item: ModuleInstallPlanItem, kind: str) -> str | None:
+    metadata = _artifact_metadata(item.artifact_integrity, kind)
+    path = _artifact_path(metadata) if metadata is not None else None
+    if path is None:
+        return None
+    return path.as_uri() if kind == "webui" else str(path)
+
+
 def _structured_webui_followup_commands(
     *,
     webui_changed: bool,
@@ -1932,9 +1989,7 @@ def _package_catalog_preflight_issues(
         return ()
     catalog_items = tuple(item for item in package_items if item.source == "catalog")
     try:
-        from govoplan_core.core.module_package_catalog import validate_module_package_catalog
-
-        result = validate_module_package_catalog()
+        result = _validate_catalog_for_plan(catalog_items)
     except Exception as exc:
         return _catalog_validation_exception_issues(exc, catalog_items=bool(catalog_items))
     issues = list(_catalog_validation_result_issues(result, catalog_items=bool(catalog_items)))
@@ -1942,10 +1997,33 @@ def _package_catalog_preflight_issues(
         return tuple(issues)
     issues.extend(_catalog_warning_issues(result))
     if catalog_items:
+        issues.extend(_catalog_plan_binding_issues(catalog_items, result))
         issues.extend(_selected_catalog_interface_issues(catalog_items, result, available))
     return tuple(issues)
 
 
+def _validate_catalog_for_plan(
+    catalog_items: tuple[ModuleInstallPlanItem, ...],
+) -> dict[str, object]:
+    from govoplan_core.core.module_package_catalog import (
+        OFFICIAL_MODULE_PACKAGE_CATALOG_URL,
+        validate_module_package_catalog,
+        validate_official_module_package_catalog,
+    )
+
+    configured = validate_module_package_catalog()
+    if configured.get("configured") or not catalog_items:
+        return configured
+    sources = {
+        str(item.catalog.get("source") or "")
+        for item in catalog_items
+        if isinstance(item.catalog, Mapping)
+    }
+    if sources == {OFFICIAL_MODULE_PACKAGE_CATALOG_URL}:
+        return validate_official_module_package_catalog()
+    return configured
+
+
 def _catalog_validation_exception_issues(exc: Exception, *, catalog_items: bool) -> tuple[ModuleInstallerIssue, ...]:
     severity: IssueSeverity = "blocker" if catalog_items else "warning"
     return (ModuleInstallerIssue(
@@ -1985,6 +2063,97 @@ def _catalog_warning_issues(result: Mapping[str, object]) -> tuple[ModuleInstall
     return tuple(ModuleInstallerIssue("warning", "catalog_warning", str(warning)) for warning in warnings)
 
 
+def _catalog_plan_binding_issues(
+    items: tuple[ModuleInstallPlanItem, ...],
+    validation: Mapping[str, object],
+) -> tuple[ModuleInstallerIssue, ...]:
+    """Require every trusted plan row to match its signed catalog entry exactly."""
+
+    modules = _catalog_modules_by_id(validation)
+    issues: list[ModuleInstallerIssue] = []
+    for item in items:
+        entry = modules.get(item.module_id)
+        if entry is None or entry.get("action") not in PACKAGE_TARGET_ACTIONS:
+            issues.append(ModuleInstallerIssue(
+                "blocker",
+                "catalog_plan_entry_missing",
+                f"The validated catalog no longer contains an install or update entry for {item.module_id!r}.",
+                item.module_id,
+            ))
+            continue
+        mismatches = _catalog_plan_entry_mismatches(item, entry, validation)
+        if mismatches:
+            issues.append(ModuleInstallerIssue(
+                "blocker",
+                "catalog_plan_binding_mismatch",
+                (
+                    "The saved package plan differs from its validated signed catalog entry "
+                    f"for: {', '.join(mismatches)}. Remove and add the catalog item again."
+                ),
+                item.module_id,
+            ))
+    return tuple(issues)
+
+
+def _catalog_plan_entry_mismatches(
+    item: ModuleInstallPlanItem,
+    entry: Mapping[str, object],
+    validation: Mapping[str, object],
+) -> tuple[str, ...]:
+    mismatches: list[str] = []
+    for attribute in ("python_package", "python_ref", "webui_package", "webui_ref"):
+        if getattr(item, attribute) != _catalog_optional_string(entry, attribute):
+            mismatches.append(attribute)
+
+    if _catalog_integrity_identity(item.artifact_integrity) != _catalog_integrity_identity(
+        entry.get("artifact_integrity")
+    ):
+        mismatches.append("artifact_integrity")
+
+    catalog = item.catalog if isinstance(item.catalog, Mapping) else {}
+    expected_snapshot = {
+        "source": validation.get("source") or validation.get("path"),
+        "channel": validation.get("channel"),
+        "sequence": validation.get("sequence"),
+        "signed": bool(validation.get("signed")),
+        "trusted": bool(validation.get("trusted")),
+        "key_id": validation.get("key_id"),
+    }
+    for attribute, expected in expected_snapshot.items():
+        actual = catalog.get(attribute)
+        if actual != expected:
+            mismatches.append(f"catalog.{attribute}")
+    return tuple(mismatches)
+
+
+def _catalog_integrity_identity(value: object) -> dict[str, dict[str, object]]:
+    if not isinstance(value, Mapping):
+        return {}
+    identity: dict[str, dict[str, object]] = {}
+    for kind in ("python", "webui"):
+        raw = value.get(kind)
+        if not isinstance(raw, Mapping):
+            continue
+        identity[kind] = {
+            field: raw.get(field)
+            for field in (
+                "ref",
+                "url",
+                "filename",
+                "sha256",
+                "size",
+                "integrity",
+                "sbom_url",
+                "provenance_url",
+                "registry_identity",
+                "git_ref",
+                "source_commit",
+            )
+            if raw.get(field) is not None
+        }
+    return identity
+
+
 def _module_install_target_plan(
     plan: ModuleInstallPlan,
     available: Mapping[str, ModuleManifest],
@@ -2817,12 +2986,15 @@ def _topological_cycle_ids(incoming: Mapping[str, set[str]]) -> tuple[str, ...]:
 def _catalog_modules_for_target_plan(
     planned_items: tuple[ModuleInstallPlanItem, ...],
 ) -> dict[str, Mapping[str, object]]:
-    if not any(item.source == "catalog" and item.action in PACKAGE_TARGET_ACTIONS for item in planned_items):
+    catalog_items = tuple(
+        item
+        for item in planned_items
+        if item.source == "catalog" and item.action in PACKAGE_TARGET_ACTIONS
+    )
+    if not catalog_items:
         return {}
     try:
-        from govoplan_core.core.module_package_catalog import validate_module_package_catalog
-
-        result = validate_module_package_catalog()
+        result = _validate_catalog_for_plan(catalog_items)
     except Exception:
         return {}
     if result.get("valid") is not True:
@@ -3772,6 +3944,156 @@ def _configured_require_artifact_integrity() -> bool:
     return os.getenv("GOVOPLAN_MODULE_INSTALLER_REQUIRE_ARTIFACT_INTEGRITY", "").strip().lower() in {"1", "true", "yes", "on"}
 
 
+def acquire_catalog_package_artifacts(
+    plan: ModuleInstallPlan,
+    *,
+    runtime_dir: Path,
+) -> ModuleInstallPlan:
+    """Materialize trusted catalog archives before package mutation."""
+
+    items: list[ModuleInstallPlanItem] = []
+    for item in plan.items:
+        if item.status != "planned" or item.action not in PACKAGE_TARGET_ACTIONS:
+            items.append(item)
+            continue
+        raw_integrity = item.artifact_integrity
+        if not isinstance(raw_integrity, Mapping):
+            items.append(item)
+            continue
+        integrity: dict[str, object] = dict(raw_integrity)
+        changed = False
+        for kind in ("python", "webui"):
+            metadata = _artifact_metadata(integrity, kind)
+            if metadata is None or _artifact_path(metadata) is not None:
+                continue
+            if not _catalog_artifact_acquisition_ready(item, metadata):
+                continue
+            updated = dict(metadata)
+            updated["artifact_path"] = str(
+                _acquire_package_artifact(
+                    metadata,
+                    runtime_dir=runtime_dir,
+                    module_id=item.module_id,
+                    kind=kind,
+                )
+            )
+            integrity[kind] = updated
+            changed = True
+        items.append(replace(item, artifact_integrity=integrity) if changed else item)
+    return replace(plan, items=tuple(items))
+
+
+def _catalog_artifact_acquisition_ready(
+    item: ModuleInstallPlanItem,
+    metadata: Mapping[str, object],
+) -> bool:
+    catalog = item.catalog
+    if (
+        item.source != "catalog"
+        or not isinstance(catalog, Mapping)
+        or catalog.get("signed") is not True
+        or catalog.get("trusted") is not True
+    ):
+        return False
+    url = _artifact_text(metadata, "url")
+    filename = _artifact_text(metadata, "filename")
+    sha256 = _artifact_text(metadata, "sha256")
+    size = metadata.get("size")
+    return bool(
+        url
+        and url.startswith("https://")
+        and filename
+        and Path(filename).name == filename
+        and sha256
+        and re.fullmatch(r"[0-9a-f]{64}", sha256.lower())
+        and isinstance(size, int)
+        and not isinstance(size, bool)
+        and 0 < size <= MAX_PACKAGE_ARTIFACT_BYTES
+    )
+
+
+def _acquire_package_artifact(
+    metadata: Mapping[str, object],
+    *,
+    runtime_dir: Path,
+    module_id: str,
+    kind: str,
+) -> Path:
+    url = validate_http_url(_artifact_text(metadata, "url") or "", label=f"{kind.capitalize()} package URL")
+    if not url.startswith("https://"):
+        raise ModuleInstallerError(f"{kind.capitalize()} package URL must use HTTPS.")
+    filename = _artifact_text(metadata, "filename") or ""
+    expected_sha256 = (_artifact_text(metadata, "sha256") or "").lower()
+    expected_size = metadata.get("size")
+    if (
+        not filename
+        or Path(filename).name != filename
+        or re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9._+!-]{0,255}", filename) is None
+        or re.fullmatch(r"[0-9a-f]{64}", expected_sha256) is None
+        or not isinstance(expected_size, int)
+        or isinstance(expected_size, bool)
+        or not 0 < expected_size <= MAX_PACKAGE_ARTIFACT_BYTES
+    ):
+        raise ModuleInstallerError(f"Catalog artifact metadata is incomplete for {module_id}/{kind}.")
+
+    cache_root = runtime_dir / "artifacts"
+    _ensure_private_artifact_directory(cache_root)
+    digest_root = cache_root / expected_sha256
+    _ensure_private_artifact_directory(digest_root)
+    target = digest_root / filename
+    if target.exists() or target.is_symlink():
+        if target.is_symlink() or not target.is_file():
+            raise ModuleInstallerError(f"Cached package artifact is not a regular file: {target}")
+        if target.stat().st_size != expected_size or _sha256_file(target) != expected_sha256:
+            raise ModuleInstallerError(f"Cached package artifact does not match its catalog identity: {target}")
+        return target
+
+    try:
+        response = fetch_http(
+            url,
+            timeout=float(os.getenv("GOVOPLAN_MODULE_INSTALLER_DOWNLOAD_TIMEOUT_SECONDS", "120")),
+            label=f"{module_id} {kind} package URL",
+            max_bytes=min(expected_size + 1, MAX_PACKAGE_ARTIFACT_BYTES),
+        )
+    except (OSError, ValueError) as exc:
+        raise ModuleInstallerError(f"Could not download {module_id} {kind} package: {exc}") from exc
+    if response.status < 200 or response.status >= 300:
+        raise ModuleInstallerError(f"Could not download {module_id} {kind} package: HTTP {response.status}.")
+    if len(response.body) != expected_size or hashlib.sha256(response.body).hexdigest() != expected_sha256:
+        raise ModuleInstallerError(f"Downloaded {module_id} {kind} package does not match its signed catalog identity.")
+
+    temporary_path: Path | None = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            mode="wb",
+            prefix=f".{filename}.",
+            suffix=".tmp",
+            dir=digest_root,
+            delete=False,
+        ) as handle:
+            temporary_path = Path(handle.name)
+            handle.write(response.body)
+            handle.flush()
+            os.fsync(handle.fileno())
+        temporary_path.chmod(0o600)
+        os.replace(temporary_path, target)
+        target.chmod(0o600)
+    except OSError as exc:
+        if temporary_path is not None:
+            temporary_path.unlink(missing_ok=True)
+        raise ModuleInstallerError(f"Could not cache {module_id} {kind} package.") from exc
+    return target
+
+
+def _ensure_private_artifact_directory(path: Path) -> None:
+    if path.is_symlink():
+        raise ModuleInstallerError(f"Installer artifact cache must not be a symlink: {path}")
+    path.mkdir(parents=True, mode=0o700, exist_ok=True)
+    path.chmod(0o700)
+    if not path.is_dir() or stat.S_IMODE(path.stat().st_mode) != 0o700:
+        raise ModuleInstallerError(f"Installer artifact cache is not private: {path}")
+
+
 def _verify_artifact_integrity(
     planned_items: tuple[ModuleInstallPlanItem, ...],
     *,
@@ -3839,7 +4161,17 @@ def _verify_artifact_metadata(
     }
     if package_name:
         record["package"] = package_name
-    for key in ("sha256", "sbom_url", "provenance_url", "registry_identity", "git_ref"):
+    for key in (
+        "sha256",
+        "url",
+        "filename",
+        "integrity",
+        "sbom_url",
+        "provenance_url",
+        "registry_identity",
+        "git_ref",
+        "source_commit",
+    ):
         value = _artifact_text(metadata, key)
         if value:
             record[key] = value
@@ -3860,6 +4192,15 @@ def _verify_artifact_metadata(
             item.module_id,
         ))
         return record, tuple(issues)
+    if artifact_path is None and _catalog_artifact_acquisition_ready(item, metadata):
+        record["acquisition_pending"] = True
+        issues.append(ModuleInstallerIssue(
+            "info",
+            "artifact_acquisition_pending",
+            f"{kind.capitalize()} artifact will be downloaded and verified by the installer daemon before package mutation.",
+            item.module_id,
+        ))
+        return record, tuple(issues)
     if artifact_path is None:
         issues.append(ModuleInstallerIssue(
             "blocker" if require_verified else "warning",

src/govoplan_core/core/module_package_catalog.py

@@ -6,6 +6,7 @@ from collections import defaultdict
 from collections.abc import Mapping
 from dataclasses import dataclass
 from datetime import UTC, datetime
+from importlib.resources import files
 from pathlib import Path
 import json
 import os
@@ -17,6 +18,10 @@ from cryptography.hazmat.primitives import serialization
 from cryptography.hazmat.primitives.asymmetric.ed25519 import Ed25519PrivateKey, Ed25519PublicKey
 
 from govoplan_core.core.versioning import format_version_range, version_range_is_valid, version_satisfies_range
+from govoplan_core.core.information_governance import (
+    information_governance_from_mapping,
+    information_governance_maturity_issues,
+)
 from govoplan_core.core.provider_governance import (
     external_provider_from_mapping,
     module_architecture_from_mapping,
@@ -25,6 +30,11 @@ from govoplan_core.core.provider_governance import (
 from govoplan_core.security.http_fetch import fetch_http_text, is_http_url
 
 _INTERFACE_NAME_RE = re.compile(r"^[a-z][a-z0-9_]*(?:\.[a-z][a-z0-9_]*)+$")
+_SHA256_RE = re.compile(r"^[0-9a-f]{64}$")
+_ARTIFACT_FILENAME_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._+!-]{0,255}$")
+_SOURCE_REPOSITORY_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._/-]{0,254}$")
+_SOURCE_REF_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._/+!-]{0,127}$")
+_SOURCE_COMMIT_RE = re.compile(r"^(?:[0-9a-f]{40}|[0-9a-f]{64})$")
 CATALOG_MIGRATION_SAFETY = ("automatic", "requires_review", "forward_only", "destructive")
 CATALOG_MIGRATION_TASK_PHASES = (
     "pre_migration_check",
@@ -32,6 +42,8 @@ CATALOG_MIGRATION_TASK_PHASES = (
     "post_migration_backfill",
     "post_migration_verify",
 )
+OFFICIAL_MODULE_PACKAGE_CATALOG_URL = "https://govoplan.add-ideas.de/catalogs/v1/channels/stable.json"
+OFFICIAL_MODULE_PACKAGE_CATALOG_CHANNEL = "stable"
 
 
 @dataclass(frozen=True, slots=True)
@@ -97,6 +109,18 @@ def validate_module_package_catalog(
     return _valid_catalog_result(catalog_source, state)
 
 
+def validate_official_module_package_catalog() -> dict[str, object]:
+    """Read the public GovOPlaN directory against Core's pinned trust anchor."""
+
+    keyring = files("govoplan_core").joinpath("resources/catalog-keyring.json").read_text(encoding="utf-8")
+    return validate_module_package_catalog(
+        OFFICIAL_MODULE_PACKAGE_CATALOG_URL,
+        require_trusted=True,
+        approved_channels=(OFFICIAL_MODULE_PACKAGE_CATALOG_CHANNEL,),
+        trusted_keys=_parse_trusted_keys(keyring),
+    )
+
+
 def _catalog_validation_state(
     source: Path | str | None,
     *,
@@ -319,7 +343,10 @@ def _configured_require_signature() -> bool:
 
 
 def _configured_approved_channels() -> tuple[str, ...]:
-    value = os.getenv("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS", "").strip()
+    value = (
+        os.getenv("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNELS", "").strip()
+        or os.getenv("GOVOPLAN_MODULE_PACKAGE_CATALOG_APPROVED_CHANNEL", "").strip()
+    )
     if not value:
         return ()
     return tuple(item.strip() for item in value.split(",") if item.strip())
@@ -628,8 +655,29 @@ def _normalize_catalog_item(value: Any) -> dict[str, object]:
         "requires_interfaces": _normalize_catalog_interface_requirements(value.get("requires_interfaces"), module_id=module_id),
         "notes": _optional_str(value, "notes"),
         "tags": _string_list(value.get("tags")),
+        "availability": _catalog_availability(value, module_id=module_id),
+        "availability_reason": _optional_str(value, "availability_reason"),
+        "configuration_requirements": _string_list(value.get("configuration_requirements")),
+        "permissions": _normalize_catalog_permissions(
+            value.get("permissions"),
+            module_id=module_id,
+        ),
     }
+    if item["availability"] == "withdrawn" and not item["availability_reason"]:
+        raise ValueError(
+            f"Withdrawn module package catalog entry {module_id!r} requires availability_reason."
+        )
+    release_notes_url = _optional_str(value, "release_notes_url")
+    if release_notes_url is not None:
+        item["release_notes_url"] = _catalog_https_url(
+            release_notes_url,
+            label=f"Module package catalog release_notes_url for {module_id!r}",
+        )
+    source = _normalize_catalog_source(value.get("source"), module_id=module_id)
+    if source:
+        item["source"] = source
     raw_architecture = value.get("architecture")
+    architecture_maturity: str | None = None
     if raw_architecture is not None:
         if not isinstance(raw_architecture, Mapping):
             raise ValueError(
@@ -647,6 +695,27 @@ def _normalize_catalog_item(value: Any) -> dict[str, object]:
                 + "; ".join(issues)
             )
         item["architecture"] = architecture.to_dict()
+        architecture_maturity = architecture.maturity
+    raw_information_governance = value.get("information_governance")
+    if raw_information_governance is not None:
+        if not isinstance(raw_information_governance, Mapping):
+            raise ValueError(
+                "Module package catalog information_governance for "
+                f"{module_id!r} must be an object."
+            )
+        information_governance = information_governance_from_mapping(
+            raw_information_governance
+        )
+        governance_issues = information_governance_maturity_issues(
+            information_governance,
+            maturity=architecture_maturity,
+        )
+        if governance_issues:
+            raise ValueError(
+                "Module package catalog information_governance for "
+                f"{module_id!r} is invalid: " + "; ".join(governance_issues)
+            )
+        item["information_governance"] = information_governance.to_dict()
     raw_providers = value.get("external_providers")
     if raw_providers is not None:
         if not isinstance(raw_providers, list):
@@ -710,6 +779,170 @@ def _normalize_catalog_item(value: Any) -> dict[str, object]:
     return item
 
 
+def _catalog_availability(value: Mapping[str, object], *, module_id: str) -> str:
+    availability = str(value.get("availability") or "available").strip().lower()
+    if availability not in {"available", "withdrawn"}:
+        raise ValueError(
+            f"Unsupported catalog availability for {module_id!r}: {availability!r}."
+        )
+    return availability
+
+
+def _normalize_catalog_permissions(
+    value: object,
+    *,
+    module_id: str,
+) -> list[dict[str, object]]:
+    if value is None:
+        return []
+    if not isinstance(value, list):
+        raise ValueError(
+            f"Module package catalog permissions for {module_id!r} must be a list."
+        )
+    if len(value) > 1000:
+        raise ValueError(
+            f"Module package catalog permissions for {module_id!r} exceed 1000 entries."
+        )
+    normalized: list[dict[str, object]] = []
+    seen: set[str] = set()
+    for raw in value:
+        if not isinstance(raw, Mapping):
+            raise ValueError(
+                f"Module package catalog permission entries for {module_id!r} must be objects."
+            )
+        scope = _bounded_catalog_permission_text(
+            raw,
+            "scope",
+            module_id=module_id,
+            maximum=200,
+        )
+        if scope in seen:
+            raise ValueError(
+                f"Module package catalog entry {module_id!r} declares permission {scope!r} more than once."
+            )
+        seen.add(scope)
+        level = _bounded_catalog_permission_text(
+            raw,
+            "level",
+            module_id=module_id,
+            maximum=20,
+        )
+        if level not in {"system", "tenant"}:
+            raise ValueError(
+                f"Module package catalog permission {module_id!r}/{scope!r} has unsupported level {level!r}."
+            )
+        deprecated = raw.get("deprecated", False)
+        if not isinstance(deprecated, bool):
+            raise ValueError(
+                f"Module package catalog permission {module_id!r}/{scope!r} deprecated must be true or false."
+            )
+        normalized.append(
+            {
+                "scope": scope,
+                "label": _bounded_catalog_permission_text(
+                    raw,
+                    "label",
+                    module_id=module_id,
+                    maximum=200,
+                ),
+                "description": _bounded_catalog_permission_text(
+                    raw,
+                    "description",
+                    module_id=module_id,
+                    maximum=1000,
+                ),
+                "category": _bounded_catalog_permission_text(
+                    raw,
+                    "category",
+                    module_id=module_id,
+                    maximum=120,
+                ),
+                "level": level,
+                "resource": _bounded_catalog_permission_text(
+                    raw,
+                    "resource",
+                    module_id=module_id,
+                    maximum=120,
+                ),
+                "action": _bounded_catalog_permission_text(
+                    raw,
+                    "action",
+                    module_id=module_id,
+                    maximum=120,
+                ),
+                "deprecated": deprecated,
+            }
+        )
+    return normalized
+
+
+def _bounded_catalog_permission_text(
+    value: Mapping[str, object],
+    key: str,
+    *,
+    module_id: str,
+    maximum: int,
+) -> str:
+    text = _required_str(value, key)
+    if len(text) > maximum:
+        raise ValueError(
+            f"Module package catalog permission {key!r} for {module_id!r} exceeds {maximum} characters."
+        )
+    return text
+
+
+def _normalize_catalog_source(
+    value: object,
+    *,
+    module_id: str,
+) -> dict[str, object]:
+    if value is None:
+        return {}
+    if not isinstance(value, Mapping):
+        raise ValueError(
+            f"Module package catalog source for {module_id!r} must be an object."
+        )
+    repository = _required_str(value, "repository")
+    tag = _required_str(value, "tag")
+    commit = _required_str(value, "commit").lower()
+    if (
+        _SOURCE_REPOSITORY_RE.fullmatch(repository) is None
+        or repository.startswith("/")
+        or repository.endswith("/")
+        or ".." in repository.split("/")
+    ):
+        raise ValueError(
+            f"Module package catalog source repository for {module_id!r} is invalid."
+        )
+    if _SOURCE_REF_RE.fullmatch(tag) is None or ".." in tag.split("/"):
+        raise ValueError(
+            f"Module package catalog source tag for {module_id!r} is invalid."
+        )
+    if _SOURCE_COMMIT_RE.fullmatch(commit) is None:
+        raise ValueError(
+            f"Module package catalog source commit for {module_id!r} is invalid."
+        )
+    source: dict[str, object] = {
+        "repository": repository,
+        "tag": tag,
+        "commit": commit,
+    }
+    for field in ("repository_url", "revision_url"):
+        url = _optional_str(value, field)
+        if url is not None:
+            source[field] = _catalog_https_url(
+                url,
+                label=f"Module package catalog source {field} for {module_id!r}",
+            )
+    return source
+
+
+def _catalog_https_url(value: str, *, label: str) -> str:
+    if not is_http_url(value) or not value.startswith("https://"):
+        raise ValueError(f"{label} must use HTTPS.")
+    return value
+
+
 def _catalog_migration_safety(value: Any, *, module_id: str) -> str:
     if value is None:
         return "automatic"
@@ -780,14 +1013,14 @@ def _catalog_optional_positive_int(value: dict[str, Any], key: str, *, module_id
     return integer
 
 
-def _required_str(value: dict[str, Any], key: str) -> str:
+def _required_str(value: Mapping[str, Any], key: str) -> str:
     item = _optional_str(value, key)
     if not item:
         raise ValueError(f"Module package catalog entry is missing {key!r}.")
     return item
 
 
-def _optional_str(value: dict[str, Any], key: str) -> str | None:
+def _optional_str(value: Mapping[str, Any], key: str) -> str | None:
     item = value.get(key)
     if item is None:
         return None
@@ -943,20 +1176,40 @@ def _normalize_artifact_integrity(value: Any) -> dict[str, object]:
             continue
         if not isinstance(raw, dict):
             raise ValueError(f"Module package catalog artifact_integrity.{key} must be an object.")
-        clean = {
+        clean: dict[str, object] = {
             field: text
             for field in (
                 "ref",
                 "path",
                 "artifact_path",
+                "url",
+                "filename",
                 "sha256",
+                "integrity",
                 "sbom_url",
                 "provenance_url",
                 "registry_identity",
                 "git_ref",
+                "source_commit",
             )
             if (text := _optional_str(raw, field))
         }
+        url = clean.get("url")
+        if isinstance(url, str) and (not is_http_url(url) or not url.startswith("https://")):
+            raise ValueError(f"Module package catalog artifact_integrity.{key}.url must use HTTPS.")
+        filename = clean.get("filename")
+        if isinstance(filename, str) and _ARTIFACT_FILENAME_RE.fullmatch(filename) is None:
+            raise ValueError(f"Module package catalog artifact_integrity.{key}.filename is invalid.")
+        sha256 = clean.get("sha256")
+        if isinstance(sha256, str) and _SHA256_RE.fullmatch(sha256.lower()) is None:
+            raise ValueError(f"Module package catalog artifact_integrity.{key}.sha256 is invalid.")
+        if isinstance(sha256, str):
+            clean["sha256"] = sha256.lower()
+        size = raw.get("size")
+        if size is not None:
+            if not isinstance(size, int) or isinstance(size, bool) or size <= 0 or size > 512 * 1024 * 1024:
+                raise ValueError(f"Module package catalog artifact_integrity.{key}.size is invalid.")
+            clean["size"] = size
         if clean:
             normalized[key] = clean
     return normalized

src/govoplan_core/core/modules.py

@@ -1,9 +1,10 @@
 from __future__ import annotations
 
 from collections.abc import Callable, Iterable, Mapping, Sequence
-from dataclasses import dataclass, field
+from dataclasses import dataclass, field, replace
 from typing import Any, Literal, Protocol, TYPE_CHECKING
 
+from govoplan_core.core.information_governance import ModuleInformationGovernance
 from govoplan_core.core.ownership import OwnershipProviderRegistration
 from govoplan_core.core.provider_governance import (
     ExternalProviderDeclaration,
@@ -14,16 +15,22 @@ from govoplan_core.core.views import ViewSurface
 
 if TYPE_CHECKING:
     from fastapi import APIRouter
-    from govoplan_core.core.operations import OperationalCheckProviderRegistration
+    from govoplan_core.core.operations import (
+        OperationalCheckProviderRegistration,
+        RuntimeWorkStatusProviderRegistration,
+    )
     from govoplan_core.core.search import (
         SearchProviderRegistration,
         SearchSourceProviderRegistration,
     )
+    from govoplan_core.core.tasks import WorkItemProviderRegistration
     from govoplan_core.core.workflows import WorkflowDefinitionContribution
 
 
 SUPPORTED_MANIFEST_CONTRACT_VERSION = "1"
 SUPPORTED_FRONTEND_ASSET_MANIFEST_CONTRACT_VERSION = "1"
+SUPPORTED_PRESENTATION_CONTRACT_VERSION = "1"
+SUPPORTED_PRODUCT_SURFACE_CONTRACT_VERSION = "1"
 
 PermissionLevel = Literal["system", "tenant"]
 SubjectType = Literal["account", "membership", "group", "service_account", "tenant"]
@@ -33,7 +40,9 @@ MigrationTaskPhase = Literal[
     "post_migration_backfill",
     "post_migration_verify",
 ]
-MigrationTaskSafety = Literal["automatic", "requires_review", "forward_only", "destructive"]
+MigrationTaskSafety = Literal[
+    "automatic", "requires_review", "forward_only", "destructive"
+]
 MigrationTaskStatus = Literal["ok", "warning", "blocked", "skipped"]
 
 
@@ -74,8 +83,6 @@ class NavItem:
     surface_id: str | None = None
 
 
-
-
 @dataclass(frozen=True, slots=True)
 class FrontendRoute:
     path: str
@@ -95,6 +102,92 @@ class PublicFrontendRoute:
     order: int = 100
 
 
+@dataclass(frozen=True, slots=True)
+class ProductAreaContribution:
+    """Assign module-owned surfaces to a user-facing product area."""
+
+    id: str
+    module_id: str
+    label: str
+    icon: str
+    surface_ids: tuple[str, ...]
+    description: str | None = None
+    order: int = 100
+
+
+ProductSurfacePresentation = Literal["task", "reader", "admin", "operator"]
+ProductAvailabilityReason = Literal[
+    "authorization",
+    "policy",
+    "configuration",
+    "disabled",
+    "capability",
+    "offline",
+    "provider_degraded",
+]
+
+
+@dataclass(frozen=True, slots=True)
+class ProductAvailabilityExplanation:
+    """Explain a product outcome without making package topology user-facing."""
+
+    reason: ProductAvailabilityReason
+    title: str
+    description: str
+    resolution: str
+    responsible_role: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class ProductSurfaceContribution:
+    """Bind an owner route to a stable, cross-module product identity."""
+
+    id: str
+    module_id: str
+    label: str
+    icon: str
+    entry_path: str
+    route_path: str
+    surface_ids: tuple[str, ...]
+    unavailable: ProductAvailabilityExplanation
+    description: str | None = None
+    degraded: ProductAvailabilityExplanation | None = None
+    presentations: tuple[ProductSurfacePresentation, ...] = ("task",)
+    capability_ids: tuple[str, ...] = ()
+    search_source_ids: tuple[str, ...] = ()
+    help_context_ids: tuple[str, ...] = ()
+    documentation_topic_ids: tuple[str, ...] = ()
+    required_all: tuple[str, ...] = ()
+    required_any: tuple[str, ...] = ()
+    aliases: tuple[str, ...] = ()
+    order: int = 100
+    contract_version: str = SUPPORTED_PRODUCT_SURFACE_CONTRACT_VERSION
+
+
+@dataclass(frozen=True, slots=True)
+class QuickAccessTool:
+    """Declare a versioned, bounded module-owned Quick Access tool."""
+
+    id: str
+    module_id: str
+    category_id: str
+    label: str
+    surface_id: str
+    icon: str
+    description: str | None = None
+    full_page_path: str | None = None
+    required_all: tuple[str, ...] = ()
+    required_any: tuple[str, ...] = ()
+    order: int = 100
+    default_enabled: bool = True
+    modes: tuple[str, ...] = ("browse",)
+    contract_version: str = "1"
+    availability: Literal["global", "active_object"] = "global"
+    accepted_reference_kinds: tuple[str, ...] = ()
+    returned_reference_kinds: tuple[str, ...] = ()
+    help_context_id: str | None = None
+
+
 @dataclass(frozen=True, slots=True)
 class FrontendModule:
     module_id: str
@@ -109,6 +202,9 @@ class FrontendModule:
     nav_items: tuple[NavItem, ...] = ()
     settings_routes: tuple[FrontendRoute, ...] = ()
     view_surfaces: tuple[ViewSurface, ...] = ()
+    product_areas: tuple[ProductAreaContribution, ...] = ()
+    product_surfaces: tuple[ProductSurfaceContribution, ...] = ()
+    quick_access_tools: tuple[QuickAccessTool, ...] = ()
 
 
 @dataclass(frozen=True, slots=True)
@@ -146,7 +242,9 @@ class ModuleMigrationTaskResult:
     details: Mapping[str, Any] = field(default_factory=dict)
 
 
-ModuleMigrationTaskExecutor = Callable[[ModuleMigrationTaskContext], ModuleMigrationTaskResult | None]
+ModuleMigrationTaskExecutor = Callable[
+    [ModuleMigrationTaskContext], ModuleMigrationTaskResult | None
+]
 
 
 @dataclass(frozen=True, slots=True)
@@ -203,7 +301,9 @@ class ModuleUninstallGuardResult:
     message: str
 
 
-UninstallGuardProvider = Callable[[object | None, str], Iterable[ModuleUninstallGuardResult]]
+UninstallGuardProvider = Callable[
+    [object | None, str], Iterable[ModuleUninstallGuardResult]
+]
 
 
 @dataclass(frozen=True, slots=True)
@@ -223,7 +323,9 @@ class ModuleContext:
 DocumentationLayer = Literal["always", "configured", "available", "evidence"]
 DocumentationLinkKind = Literal["runtime", "api", "repository", "wiki", "public"]
 DocumentationType = Literal["admin", "user"]
-DocumentationConfigurationState = Literal["enabled", "disabled", "inherited", "unavailable"]
+DocumentationConfigurationState = Literal[
+    "enabled", "disabled", "inherited", "unavailable"
+]
 DocumentationSourceKind = Literal[
     "manifest",
     "route",
@@ -238,6 +340,30 @@ DocumentationSourceState = Literal["configured", "disabled", "unavailable"]
 CapabilityStability = Literal["experimental", "stable", "deprecated"]
 
 
+DOCUMENTATION_STRUCTURED_TRANSLATION_VERSION = "1"
+DOCUMENTATION_LOCALIZABLE_METADATA_KEYS = frozenset(
+    {
+        "admin_explanation",
+        "consequence_classes",
+        "consequences",
+        "constraints",
+        "current_configuration",
+        "fields",
+        "limitations",
+        "operational_consequences",
+        "outcome",
+        "prerequisites",
+        "privacy_notes",
+        "purpose",
+        "result",
+        "steps",
+        "user_explanation",
+        "verification",
+        "when_used",
+    }
+)
+
+
 @dataclass(frozen=True, slots=True)
 class DocumentationLink:
     label: str
@@ -273,12 +399,161 @@ class DocumentationTopic:
     configuration_keys: tuple[str, ...] = ()
     i18n_key: str | None = None
     translations: Mapping[str, Mapping[str, str]] = field(default_factory=dict)
+    structured_translation_version: str | None = None
+    structured_translations: Mapping[str, Mapping[str, Any]] = field(
+        default_factory=dict
+    )
     source_module_id: str | None = None
     version_min: str | None = None
     version_max_exclusive: str | None = None
     metadata: Mapping[str, Any] = field(default_factory=dict)
 
 
+def localize_documentation_topics(
+    topics: Iterable[DocumentationTopic],
+    *,
+    locale: str,
+    translations: Mapping[str, Mapping[str, str]],
+) -> tuple[DocumentationTopic, ...]:
+    """Merge owner-supplied text translations without moving feature content."""
+    localized: list[DocumentationTopic] = []
+    for topic in topics:
+        translated = translations.get(topic.id)
+        if translated is None:
+            localized.append(topic)
+            continue
+        values = {name: dict(value) for name, value in topic.translations.items()}
+        values[locale] = {**values.get(locale, {}), **translated}
+        localized.append(replace(topic, translations=values))
+    return tuple(localized)
+
+
+def localizable_documentation_metadata_keys(
+    topic: DocumentationTopic,
+) -> tuple[str, ...]:
+    """Return structured metadata keys whose values are public prose."""
+
+    return tuple(
+        sorted(DOCUMENTATION_LOCALIZABLE_METADATA_KEYS.intersection(topic.metadata))
+    )
+
+
+def localized_documentation_metadata(
+    topic: DocumentationTopic,
+    locale: str,
+) -> dict[str, Any]:
+    """Overlay one validated structured translation onto source metadata."""
+
+    localized = dict(topic.metadata)
+    translation = topic.structured_translations.get(locale)
+    if translation:
+        localized.update(translation)
+    return localized
+
+
+def documentation_structured_translation_issues(
+    topic: DocumentationTopic,
+) -> tuple[str, ...]:
+    """Validate the opt-in, versioned structured-documentation translation."""
+
+    version = topic.structured_translation_version
+    translations = topic.structured_translations
+    if version is None:
+        if translations:
+            return (
+                "structured_translations require structured_translation_version",
+            )
+        return ()
+    if version != DOCUMENTATION_STRUCTURED_TRANSLATION_VERSION:
+        return (
+            "unsupported structured_translation_version "
+            f"{version!r}; expected {DOCUMENTATION_STRUCTURED_TRANSLATION_VERSION!r}",
+        )
+
+    localizable_keys = set(localizable_documentation_metadata_keys(topic))
+    issues: list[str] = []
+    for locale, translation in translations.items():
+        if not locale.strip():
+            issues.append("structured translation locale must not be empty")
+            continue
+        translated_keys = set(translation)
+        for key in sorted(translated_keys - localizable_keys):
+            issues.append(
+                f"structured translation {locale!r} contains non-localizable or missing metadata key {key!r}"
+            )
+        for key in sorted(localizable_keys - translated_keys):
+            issues.append(
+                f"structured translation {locale!r} is missing metadata key {key!r}"
+            )
+        for key in sorted(localizable_keys & translated_keys):
+            issues.extend(
+                _structured_translation_shape_issues(
+                    topic.metadata[key],
+                    translation[key],
+                    path=f"{locale}.{key}",
+                )
+            )
+    return tuple(issues)
+
+
+def _structured_translation_shape_issues(
+    source: object,
+    translated: object,
+    *,
+    path: str,
+) -> tuple[str, ...]:
+    if isinstance(source, str):
+        if not isinstance(translated, str) or not translated.strip():
+            return (f"structured translation {path} must be a non-empty string",)
+        return ()
+    if isinstance(source, Mapping):
+        if not isinstance(translated, Mapping):
+            return (f"structured translation {path} must preserve object shape",)
+        issues: list[str] = []
+        source_keys = {str(key) for key in source}
+        translated_keys = {str(key) for key in translated}
+        if source_keys != translated_keys:
+            issues.append(
+                f"structured translation {path} must preserve object keys"
+            )
+            return tuple(issues)
+        for key, value in source.items():
+            issues.extend(
+                _structured_translation_shape_issues(
+                    value,
+                    translated[key],
+                    path=f"{path}.{key}",
+                )
+            )
+        return tuple(issues)
+    if isinstance(source, Sequence) and not isinstance(
+        source, (str, bytes, bytearray)
+    ):
+        if not isinstance(translated, Sequence) or isinstance(
+            translated, (str, bytes, bytearray)
+        ):
+            return (f"structured translation {path} must preserve list shape",)
+        if len(source) != len(translated):
+            return (f"structured translation {path} must preserve list length",)
+        issues: list[str] = []
+        for index, (source_item, translated_item) in enumerate(
+            zip(source, translated, strict=True)
+        ):
+            issues.extend(
+                _structured_translation_shape_issues(
+                    source_item,
+                    translated_item,
+                    path=f"{path}[{index}]",
+                )
+            )
+        return tuple(issues)
+    if translated != source:
+        return (
+            f"structured translation {path} must preserve non-text value {source!r}",
+        )
+    return ()
+
+
 def user_workflow_scope_condition_issues(topic: DocumentationTopic) -> tuple[str, ...]:
     """Return fail-closed authoring issues for a user-facing workflow topic.
 
@@ -288,16 +563,23 @@ def user_workflow_scope_condition_issues(topic: DocumentationTopic) -> tuple[str
     """
 
     raw_kind = topic.metadata.get("kind")
-    kind = raw_kind.strip().lower().replace("_", "-") if isinstance(raw_kind, str) else ""
+    kind = (
+        raw_kind.strip().lower().replace("_", "-") if isinstance(raw_kind, str) else ""
+    )
     if kind != "workflow" or "user" not in topic.documentation_types:
         return ()
     if not topic.conditions:
-        return ("user workflow topics must declare at least one scope-conditioned alternative",)
+        return (
+            "user workflow topics must declare at least one scope-conditioned alternative",
+        )
 
     unscoped_alternatives = tuple(
         index
         for index, condition in enumerate(topic.conditions, start=1)
-        if not any(scope.strip() for scope in (*condition.required_scopes, *condition.any_scopes))
+        if not any(
+            scope.strip()
+            for scope in (*condition.required_scopes, *condition.any_scopes)
+        )
     )
     if not unscoped_alternatives:
         return ()
@@ -369,14 +651,11 @@ class CapabilityDocumentation:
 class ResourceAclProvider(Protocol):
     resource_type: str
 
-    def can_read(self, principal: object, resource_id: str) -> bool:
-        ...
+    def can_read(self, principal: object, resource_id: str) -> bool: ...
 
-    def can_write(self, principal: object, resource_id: str) -> bool:
-        ...
+    def can_write(self, principal: object, resource_id: str) -> bool: ...
 
-    def explain(self, principal: object, resource_id: str) -> AccessDecision:
-        ...
+    def explain(self, principal: object, resource_id: str) -> AccessDecision: ...
 
 
 TenantSummaryProvider = Callable[[object, str], Mapping[str, int]]
@@ -435,17 +714,29 @@ class ModuleManifest:
     ownership_providers: tuple[OwnershipProviderRegistration, ...] = ()
     tenant_summary_providers: tuple[TenantSummaryProvider, ...] = ()
     tenant_summary_batch_providers: tuple[TenantSummaryBatchProvider, ...] = ()
-    delete_veto_providers: Mapping[str, Sequence[DeleteVetoProvider]] = field(default_factory=dict)
+    delete_veto_providers: Mapping[str, Sequence[DeleteVetoProvider]] = field(
+        default_factory=dict
+    )
     uninstall_guard_providers: tuple[UninstallGuardProvider, ...] = ()
     capability_factories: Mapping[str, CapabilityFactory] = field(default_factory=dict)
-    capability_documentation: Mapping[str, CapabilityDocumentation] = field(default_factory=dict)
+    capability_documentation: Mapping[str, CapabilityDocumentation] = field(
+        default_factory=dict
+    )
     search_providers: tuple["SearchProviderRegistration", ...] = ()
     search_sources: tuple["SearchSourceProviderRegistration", ...] = ()
+    work_item_providers: tuple["WorkItemProviderRegistration", ...] = ()
     operational_check_providers: tuple[
         "OperationalCheckProviderRegistration",
         ...,
     ] = ()
+    runtime_work_status_providers: tuple[
+        "RuntimeWorkStatusProviderRegistration",
+        ...,
+    ] = ()
     architecture: ModuleArchitectureDeclaration | None = None
+    information_governance: ModuleInformationGovernance = field(
+        default_factory=ModuleInformationGovernance
+    )
     external_providers: tuple[ExternalProviderDeclaration, ...] = ()
     external_provider_state_providers: tuple[
         ExternalProviderStateProviderRegistration,
@@ -466,3 +757,53 @@ class ModuleManifest:
     # runtime module ID changes.
     permission_namespace: str | None = None
     workflow_definitions: tuple["WorkflowDefinitionContribution", ...] = ()
+
+
+def with_documentation_structured_translations(
+    manifest: ModuleManifest,
+    *,
+    locale: str,
+    translations: Mapping[str, Mapping[str, Any]],
+) -> ModuleManifest:
+    """Merge module-owned structured documentation translations by topic id.
+
+    The helper keeps feature prose in its owning module while giving every
+    manifest the same fail-closed merge behavior. Unknown topic ids and
+    incomplete or shape-changing locale maps are rejected immediately.
+    """
+
+    locale = locale.strip()
+    if not locale:
+        raise ValueError("structured documentation locale must not be empty")
+
+    topics_by_id = {topic.id: topic for topic in manifest.documentation}
+    unknown_topic_ids = sorted(set(translations) - set(topics_by_id))
+    if unknown_topic_ids:
+        raise ValueError(
+            "structured documentation translations reference unknown topic ids: "
+            + ", ".join(unknown_topic_ids)
+        )
+
+    localized_topics: list[DocumentationTopic] = []
+    for topic in manifest.documentation:
+        translation = translations.get(topic.id)
+        if translation is None:
+            localized_topics.append(topic)
+            continue
+
+        structured_translations = dict(topic.structured_translations)
+        structured_translations[locale] = translation
+        localized_topic = replace(
+            topic,
+            structured_translation_version=DOCUMENTATION_STRUCTURED_TRANSLATION_VERSION,
+            structured_translations=structured_translations,
+        )
+        issues = documentation_structured_translation_issues(localized_topic)
+        if issues:
+            raise ValueError(
+                f"invalid {locale!r} structured documentation translation for "
+                f"{topic.id!r}: {'; '.join(issues)}"
+            )
+        localized_topics.append(localized_topic)
+
+    return replace(manifest, documentation=tuple(localized_topics))

src/govoplan_core/core/navigation.py

@@ -0,0 +1,229 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any, Iterable, Mapping
+
+NAVIGATION_PREFERENCES_KEY = "navigation_preferences"
+NAVIGATION_PREFERENCES_CONTRACT_VERSION = "1"
+_MAX_ITEMS = 256
+
+
+@dataclass(frozen=True, slots=True)
+class NavigationSeparator:
+    id: str
+    label: str = ""
+
+    def as_dict(self) -> dict[str, str]:
+        return {"id": self.id, "label": self.label}
+
+
+@dataclass(frozen=True, slots=True)
+class NavigationPreferences:
+    order: tuple[str, ...] = ()
+    hidden: tuple[str, ...] = ()
+    locked: tuple[str, ...] = ()
+    # None preserves inherited grouping; an empty tuple explicitly removes it.
+    separators: tuple[NavigationSeparator, ...] | None = None
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "contract_version": NAVIGATION_PREFERENCES_CONTRACT_VERSION,
+            "order": list(self.order),
+            "hidden": list(self.hidden),
+            "locked": list(self.locked),
+            **({"separators": [item.as_dict() for item in self.separators]} if self.separators is not None else {}),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class EffectiveNavigationItem:
+    id: str
+    order: int
+    visible: bool
+    locked: bool
+    order_source: str
+    visibility_source: str
+    lock_source: str | None = None
+    section: NavigationSeparator | None = None
+    custom_layout: bool = False
+    layout_source: str = "module"
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "navigation_id": self.id,
+            "order": self.order,
+            "navigation_visible": self.visible,
+            "navigation_locked": self.locked,
+            "navigation_order_source": self.order_source,
+            "navigation_visibility_source": self.visibility_source,
+            "navigation_lock_source": self.lock_source,
+            "navigation_section": self.section.as_dict() if self.section else None,
+            "navigation_custom_layout": self.custom_layout,
+            "navigation_layout_source": self.layout_source,
+        }
+
+
+def navigation_preferences_from_settings(
+    settings: object,
+) -> NavigationPreferences | None:
+    if not isinstance(settings, Mapping):
+        return None
+    raw = settings.get(NAVIGATION_PREFERENCES_KEY)
+    if not isinstance(raw, Mapping):
+        return None
+    return navigation_preferences_from_mapping(raw)
+
+
+def navigation_preferences_from_mapping(
+    raw: Mapping[str, Any],
+) -> NavigationPreferences:
+    return NavigationPreferences(
+        order=_ids(raw.get("order")),
+        hidden=_ids(raw.get("hidden")),
+        locked=_ids(raw.get("locked")),
+        separators=_separators(raw.get("separators")),
+    )
+
+
+def update_navigation_preferences(
+    settings: object,
+    preferences: NavigationPreferences | Mapping[str, Any] | None,
+) -> dict[str, Any]:
+    updated = dict(settings) if isinstance(settings, Mapping) else {}
+    if preferences is None:
+        updated.pop(NAVIGATION_PREFERENCES_KEY, None)
+    else:
+        raw = preferences.as_dict() if isinstance(preferences, NavigationPreferences) else preferences
+        updated[NAVIGATION_PREFERENCES_KEY] = navigation_preferences_from_mapping(
+            raw
+        ).as_dict()
+    return updated
+
+
+def resolve_navigation_preferences(
+    item_ids: Iterable[str],
+    *,
+    system: NavigationPreferences | None = None,
+    tenant: NavigationPreferences | None = None,
+    user: NavigationPreferences | None = None,
+) -> dict[str, EffectiveNavigationItem]:
+    ordered = list(dict.fromkeys(_clean_id(item_id) for item_id in item_ids))
+    ordered = [item_id for item_id in ordered if item_id]
+    available = set(ordered)
+    order_source = {item_id: "module" for item_id in ordered}
+    visibility = {item_id: True for item_id in ordered}
+    visibility_source = {item_id: "module" for item_id in ordered}
+    locks: dict[str, str] = {}
+    separators: dict[str, NavigationSeparator] = {}
+    custom_layout = False
+    layout_source = "module"
+
+    for source, preferences, may_lock in (
+        ("system", system, True),
+        ("tenant", tenant, True),
+        ("user", user, False),
+    ):
+        if preferences is None:
+            continue
+        if preferences.separators is not None:
+            separators = {item.id: item for item in preferences.separators if item.id not in available}
+            ordered = [item_id for item_id in ordered if item_id in available or item_id in separators]
+            ordered.extend(item_id for item_id in separators if item_id not in ordered)
+            custom_layout = True
+            layout_source = source
+        requested_order = list(dict.fromkeys(item_id for item_id in preferences.order if item_id in available or item_id in separators))
+        if requested_order:
+            requested = set(requested_order)
+            ordered = [*requested_order, *(item_id for item_id in ordered if item_id not in requested)]
+            for item_id in requested_order:
+                order_source[item_id] = source
+
+        requested_hidden = set(preferences.hidden).intersection(available)
+        for item_id in available:
+            if item_id in locks:
+                visibility[item_id] = True
+                visibility_source[item_id] = locks[item_id]
+                continue
+            visibility[item_id] = item_id not in requested_hidden
+            visibility_source[item_id] = source
+
+        if may_lock:
+            for item_id in preferences.locked:
+                if item_id not in available:
+                    continue
+                locks[item_id] = source
+                visibility[item_id] = True
+                visibility_source[item_id] = source
+
+    result: dict[str, EffectiveNavigationItem] = {}
+    section: NavigationSeparator | None = None
+    for index, item_id in enumerate(ordered):
+        if item_id in separators:
+            section = separators[item_id]
+            continue
+        result[item_id] = EffectiveNavigationItem(
+            id=item_id,
+            order=index,
+            visible=visibility[item_id],
+            locked=item_id in locks,
+            order_source=order_source[item_id],
+            visibility_source=visibility_source[item_id],
+            lock_source=locks.get(item_id),
+            section=section,
+            custom_layout=custom_layout,
+            layout_source=layout_source,
+        )
+    return result
+
+
+def _separators(value: object) -> tuple[NavigationSeparator, ...] | None:
+    if not isinstance(value, (list, tuple)):
+        return None
+    items: dict[str, NavigationSeparator] = {}
+    for raw in value[:_MAX_ITEMS]:
+        if not isinstance(raw, Mapping):
+            continue
+        item_id = _clean_id(raw.get("id"))
+        label = raw.get("label", "")
+        if not item_id.startswith("separator:") or not isinstance(label, str):
+            continue
+        label = label.strip()[:120]
+        if any(ord(character) < 32 for character in label):
+            continue
+        items[item_id] = NavigationSeparator(item_id, label)
+    return tuple(items.values())
+
+
+def _ids(value: object) -> tuple[str, ...]:
+    if not isinstance(value, (list, tuple)):
+        return ()
+    cleaned = tuple(
+        dict.fromkeys(
+            item_id
+            for item in value[:_MAX_ITEMS]
+            if (item_id := _clean_id(item))
+        )
+    )
+    return cleaned
+
+
+def _clean_id(value: object) -> str:
+    if not isinstance(value, str):
+        return ""
+    clean = value.strip()
+    if not clean or len(clean) > 255 or any(ord(character) < 32 for character in clean):
+        return ""
+    return clean
+
+
+__all__ = [
+    "EffectiveNavigationItem",
+    "NAVIGATION_PREFERENCES_CONTRACT_VERSION",
+    "NAVIGATION_PREFERENCES_KEY",
+    "NavigationPreferences",
+    "NavigationSeparator",
+    "navigation_preferences_from_mapping",
+    "navigation_preferences_from_settings",
+    "resolve_navigation_preferences",
+    "update_navigation_preferences",
+]

src/govoplan_core/core/operations.py

@@ -1,7 +1,8 @@
 from __future__ import annotations
 
-from collections.abc import Callable, Mapping
+from collections.abc import Callable, Mapping, Sequence
 from dataclasses import dataclass, field
+from datetime import datetime
 from typing import Literal
 
 
@@ -42,3 +43,84 @@ class OperationalCheckProviderRegistration:
     provider: OperationalCheckProvider
     cache_seconds: int = 60
 
+
+RuntimeWorkState = Literal[
+    "disabled",
+    "unconfigured",
+    "starting",
+    "healthy",
+    "idle",
+    "busy",
+    "degraded",
+    "stale",
+    "unreachable",
+]
+
+
+@dataclass(frozen=True, slots=True)
+class RuntimeWorkStatusContext:
+    """Sanitized process evidence supplied to a runtime-work provider."""
+
+    profile: str
+    observed_at: datetime
+    stale_after_seconds: int
+    runtime_nodes: Sequence[Mapping[str, object]] = ()
+
+
+@dataclass(frozen=True, slots=True)
+class RuntimeWorkStatus:
+    """Bounded worker/queue status with explicit unsupported metrics."""
+
+    provider_id: str
+    label: str
+    backend: str
+    enabled: bool
+    configured: bool
+    state: RuntimeWorkState
+    detail: str
+    observed_at: datetime
+    active_workers: int | None = None
+    last_heartbeat_at: datetime | None = None
+    queue_depths: Mapping[str, int | None] = field(default_factory=dict)
+    active_work: int | None = None
+    reserved_work: int | None = None
+    failures: int | None = None
+    stale_after_seconds: int | None = None
+    guidance: str = ""
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "provider_id": self.provider_id,
+            "label": self.label,
+            "backend": self.backend,
+            "enabled": self.enabled,
+            "configured": self.configured,
+            "state": self.state,
+            "detail": self.detail,
+            "observed_at": self.observed_at.isoformat(),
+            "active_workers": self.active_workers,
+            "last_heartbeat_at": (
+                self.last_heartbeat_at.isoformat()
+                if self.last_heartbeat_at is not None
+                else None
+            ),
+            "queue_depths": dict(self.queue_depths),
+            "active_work": self.active_work,
+            "reserved_work": self.reserved_work,
+            "failures": self.failures,
+            "stale_after_seconds": self.stale_after_seconds,
+            "guidance": self.guidance,
+        }
+
+
+RuntimeWorkStatusProvider = Callable[[RuntimeWorkStatusContext], RuntimeWorkStatus]
+
+
+@dataclass(frozen=True, slots=True)
+class RuntimeWorkStatusProviderRegistration:
+    """Register one optional provider-neutral worker/queue observation."""
+
+    module_id: str
+    provider_id: str
+    provider: RuntimeWorkStatusProvider
+    cache_seconds: int = 15

src/govoplan_core/core/payments.py

@@ -0,0 +1,87 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Protocol, runtime_checkable
+
+from govoplan_core.core.institutional import EvidenceReference
+
+
+CAPABILITY_PAYMENT_REQUESTS = "payments.requests"
+
+
+@dataclass(frozen=True, slots=True)
+class PaymentRequestCommand:
+    tenant_id: str
+    source_module: str
+    source_resource_type: str
+    source_resource_id: str
+    amount_minor: int
+    currency: str
+    subject: str
+    idempotency_key: str
+    requested_at: datetime
+    requested_by_ref: str
+    due_at: datetime | None = None
+    context_refs: Mapping[str, str] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+
+@dataclass(frozen=True, slots=True)
+class ManualPaymentReconciliationCommand:
+    tenant_id: str
+    payment_id: str
+    amount_minor: int
+    currency: str
+    transaction_reference: str
+    evidence_ref: EvidenceReference
+    idempotency_key: str
+    received_at: datetime
+    recorded_at: datetime
+    recorded_by_ref: str
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+
+@runtime_checkable
+class PaymentRequestProvider(Protocol):
+    def request_payment(
+        self,
+        session: object,
+        command: PaymentRequestCommand,
+    ) -> Mapping[str, object]:
+        ...
+
+    def get_payment(
+        self,
+        session: object,
+        *,
+        tenant_id: str,
+        payment_id: str,
+    ) -> Mapping[str, object] | None:
+        ...
+
+    def reconcile_manual_payment(
+        self,
+        session: object,
+        command: ManualPaymentReconciliationCommand,
+    ) -> Mapping[str, object]:
+        ...
+
+
+def payment_request_provider(registry: object | None) -> PaymentRequestProvider | None:
+    if registry is None or not hasattr(registry, "has_capability"):
+        return None
+    if not registry.has_capability(CAPABILITY_PAYMENT_REQUESTS):
+        return None
+    capability = registry.capability(CAPABILITY_PAYMENT_REQUESTS)
+    return capability if isinstance(capability, PaymentRequestProvider) else None
+
+
+__all__ = [
+    "CAPABILITY_PAYMENT_REQUESTS",
+    "ManualPaymentReconciliationCommand",
+    "PaymentRequestCommand",
+    "PaymentRequestProvider",
+    "payment_request_provider",
+]

src/govoplan_core/core/platform_interfaces.py

@@ -21,11 +21,14 @@ PlatformInterfaceKind = Literal[
     "frontend_route",
     "navigation",
     "permission",
+    "product_area",
+    "product_surface",
     "provided_interface",
     "public_route",
     "search_provider",
     "search_source",
     "settings_route",
+    "quick_access_tool",
     "view_surface",
 ]
 
@@ -217,6 +220,80 @@ def manifest_interface_declarations(
                     },
                 )
             )
+        for area in frontend.product_areas:
+            declarations.append(
+                PlatformInterfaceDeclaration(
+                    id=f"{manifest.id}.{area.id}",
+                    module_id=manifest.id,
+                    kind="product_area",
+                    label=area.label,
+                    required_all=(),
+                    required_any=(),
+                    metadata={
+                        "area_id": area.id,
+                        "icon": area.icon,
+                        "description": area.description,
+                        "order": area.order,
+                        "surface_ids": list(area.surface_ids),
+                    },
+                )
+            )
+        for surface in frontend.product_surfaces:
+            declarations.append(
+                PlatformInterfaceDeclaration(
+                    id=f"{manifest.id}.{surface.id}",
+                    module_id=manifest.id,
+                    kind="product_surface",
+                    label=surface.label,
+                    path=surface.route_path,
+                    required_all=surface.required_all,
+                    required_any=surface.required_any,
+                    metadata={
+                        "contract_version": surface.contract_version,
+                        "product_surface_id": surface.id,
+                        "description": surface.description,
+                        "icon": surface.icon,
+                        "entry_path": surface.entry_path,
+                        "surface_ids": list(surface.surface_ids),
+                        "presentations": list(surface.presentations),
+                        "capability_ids": list(surface.capability_ids),
+                        "search_source_ids": list(surface.search_source_ids),
+                        "help_context_ids": list(surface.help_context_ids),
+                        "documentation_topic_ids": list(
+                            surface.documentation_topic_ids
+                        ),
+                        "aliases": list(surface.aliases),
+                        "order": surface.order,
+                        "unavailable_reason": surface.unavailable.reason,
+                        "degraded_reason": (
+                            surface.degraded.reason
+                            if surface.degraded is not None
+                            else None
+                        ),
+                    },
+                )
+            )
+        for tool in frontend.quick_access_tools:
+            declarations.append(
+                PlatformInterfaceDeclaration(
+                    id=tool.id,
+                    module_id=manifest.id,
+                    kind="quick_access_tool",
+                    label=tool.label,
+                    path=tool.full_page_path,
+                    required_all=tool.required_all,
+                    required_any=tool.required_any,
+                    metadata={
+                        "category_id": tool.category_id,
+                        "surface_id": tool.surface_id,
+                        "icon": tool.icon,
+                        "description": tool.description,
+                        "order": tool.order,
+                        "default_enabled": tool.default_enabled,
+                        "modes": list(tool.modes),
+                    },
+                )
+            )
 
     frontend_navigation = {
         declaration.id: declaration

src/govoplan_core/core/policy.py

@@ -7,6 +7,20 @@ from urllib.parse import quote, unquote
 from govoplan_core.core.access import PrincipalRef
 
 PolicyScopeType = Literal["system", "tenant", "user", "group", "campaign"]
+PolicyImpactPopulationState = Literal[
+    "complete",
+    "sampled",
+    "truncated",
+    "unavailable",
+]
+CampaignArchiveEncryptionMethod = Literal["aes", "zip_standard"]
+CampaignArchivePasswordDeliveryChannel = Literal[
+    "separate_mail",
+    "sms",
+    "letter",
+    "phone",
+    "in_person",
+]
 SchedulingParticipantVisibility = Literal["aggregates_only", "names_and_statuses"]
 DefinitionScopeType = Literal["system", "tenant", "group", "user"]
 DefinitionKind = Literal["flow", "template"]
@@ -27,10 +41,12 @@ ViewGovernanceAction = Literal[
     "workflow_activate",
 ]
 FunctionAssignmentChangeKind = Literal["request", "grant"]
+FunctionAssignmentReviewStep = Literal["holder", "authority", "recipient"]
 FunctionAssignmentGovernanceAction = Literal[
     "submit",
     "approve_holder",
     "approve_authority",
+    "approve_escalation",
     "accept_recipient",
     "request_changes",
     "respond",
@@ -45,6 +61,8 @@ CAPABILITY_POLICY_SCHEDULING_PARTICIPANT_PRIVACY = "policy.schedulingParticipant
 CAPABILITY_POLICY_DEFINITION_GOVERNANCE = "policy.definitionGovernance"
 CAPABILITY_POLICY_VIEW_GOVERNANCE = "policy.viewGovernance"
 CAPABILITY_POLICY_FUNCTION_ASSIGNMENT_GOVERNANCE = "policy.functionAssignmentGovernance"
+CAPABILITY_POLICY_CAMPAIGN_ARCHIVE_ENCRYPTION = "policy.campaignArchiveEncryption"
+CAPABILITY_POLICY_IMPACT_SUBJECT_PREFIX = "policy.impactSubjects."
 
 POLICY_SCOPE_TYPES: tuple[PolicyScopeType, ...] = (
     "system",
@@ -182,6 +200,213 @@ class PolicyDecision:
         }
 
 
+@dataclass(frozen=True, slots=True)
+class PolicyImpactPopulationRequest:
+    """One explicit, bounded request to an optional impact-subject provider."""
+
+    tenant_id: str
+    policy_family: str
+    selector: Mapping[str, Any] = field(default_factory=dict)
+    limit: int = 200
+    actor_scopes: tuple[str, ...] = ()
+    allow_sensitive_details: bool = False
+
+    def __post_init__(self) -> None:
+        if not self.tenant_id.strip():
+            raise ValueError("Policy impact population requires a tenant ID")
+        if not self.policy_family.strip() or len(self.policy_family) > 120:
+            raise ValueError(
+                "Policy impact population family must contain 1 to 120 characters"
+            )
+        if self.limit < 1 or self.limit > 500:
+            raise ValueError("Policy impact population limit must be between 1 and 500")
+
+
+@dataclass(frozen=True, slots=True)
+class PolicyImpactSubject:
+    """Provider-owned reference safe for Policy to compare without domain imports."""
+
+    module_id: str
+    resource_type: str
+    resource_id: str
+    action: str
+    label: str | None = None
+    scope_type: PolicyScopeType | None = None
+    scope_id: str | None = None
+    attributes: Mapping[str, Any] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        for label, value, maximum in (
+            ("module ID", self.module_id, 80),
+            ("resource type", self.resource_type, 80),
+            ("resource ID", self.resource_id, 240),
+            ("action", self.action, 120),
+        ):
+            if not value.strip() or len(value) > maximum:
+                raise ValueError(
+                    f"Policy impact subject {label} must contain 1 to {maximum} characters"
+                )
+
+    @property
+    def key(self) -> tuple[str, str, str, str]:
+        return (
+            self.module_id,
+            self.resource_type,
+            self.resource_id,
+            self.action,
+        )
+
+    def to_dict(self) -> dict[str, Any]:
+        return {
+            "module_id": self.module_id,
+            "resource_type": self.resource_type,
+            "resource_id": self.resource_id,
+            "action": self.action,
+            "label": self.label,
+            "scope_type": self.scope_type,
+            "scope_id": self.scope_id,
+            "attributes": dict(self.attributes),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class PolicyImpactSubjectBatch:
+    provider_id: str
+    subjects: tuple[PolicyImpactSubject, ...] = ()
+    state: PolicyImpactPopulationState = "complete"
+    total_available: int | None = None
+    explanation: str | None = None
+
+    def __post_init__(self) -> None:
+        if not self.provider_id.strip() or len(self.provider_id) > 120:
+            raise ValueError(
+                "Policy impact provider ID must contain 1 to 120 characters"
+            )
+        if len(self.subjects) > 500:
+            raise ValueError("Policy impact providers may return at most 500 subjects")
+        if len({subject.key for subject in self.subjects}) != len(self.subjects):
+            raise ValueError("Policy impact provider returned duplicate subjects")
+        if self.total_available is not None and self.total_available < len(self.subjects):
+            raise ValueError(
+                "Policy impact population total cannot be smaller than its subjects"
+            )
+        if self.state == "unavailable" and not self.explanation:
+            raise ValueError("Unavailable policy impact populations need an explanation")
+
+    def to_dict(self, *, include_subjects: bool = True) -> dict[str, Any]:
+        return {
+            "provider_id": self.provider_id,
+            "state": self.state,
+            "returned": len(self.subjects),
+            "total_available": self.total_available,
+            "explanation": self.explanation,
+            "subjects": (
+                [subject.to_dict() for subject in self.subjects]
+                if include_subjects
+                else []
+            ),
+        }
+
+
+@runtime_checkable
+class PolicyImpactSubjectProvider(Protocol):
+    provider_id: str
+    supported_policy_families: tuple[str, ...]
+
+    def collect_policy_impact_subjects(
+        self,
+        session: object | None = None,
+        *,
+        request: PolicyImpactPopulationRequest,
+    ) -> PolicyImpactSubjectBatch: ...
+
+
+def policy_impact_subject_provider(
+    registry: object | None,
+    provider_id: str,
+) -> PolicyImpactSubjectProvider | None:
+    clean_provider_id = provider_id.strip()
+    if not clean_provider_id or registry is None:
+        return None
+    capability_name = f"{CAPABILITY_POLICY_IMPACT_SUBJECT_PREFIX}{clean_provider_id}"
+    if (
+        not hasattr(registry, "has_capability")
+        or not registry.has_capability(capability_name)
+    ):
+        return None
+    capability = registry.capability(capability_name)
+    if not isinstance(capability, PolicyImpactSubjectProvider):
+        return None
+    return capability
+
+
+@dataclass(frozen=True, slots=True)
+class CampaignArchiveEncryptionRequest:
+    """Context required to resolve one Campaign archive-encryption ceiling.
+
+    The owning module supplies the stable Campaign and owner references. Policy
+    owns hierarchy evaluation; Campaign owns archive configuration and evidence.
+    """
+
+    tenant_id: str
+    campaign_id: str
+    owner_type: Literal["user", "group"] | None = None
+    owner_id: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class CampaignArchiveEncryptionDecision:
+    allowed_password_encryption_methods: frozenset[CampaignArchiveEncryptionMethod]
+    allowed_password_delivery_channels: frozenset[
+        CampaignArchivePasswordDeliveryChannel
+    ]
+    policy_hash: str
+    source_path: tuple[PolicySourceStep, ...] = ()
+    reason: str | None = None
+    diagnostics: tuple[Mapping[str, Any], ...] = ()
+
+    def to_dict(self) -> dict[str, Any]:
+        return {
+            "allowed_password_encryption_methods": sorted(
+                self.allowed_password_encryption_methods
+            ),
+            "allowed_password_delivery_channels": sorted(
+                self.allowed_password_delivery_channels
+            ),
+            "policy_hash": self.policy_hash,
+            "source_path": [step.to_dict() for step in self.source_path],
+            "reason": self.reason,
+            "diagnostics": [dict(item) for item in self.diagnostics],
+        }
+
+
+@runtime_checkable
+class CampaignArchiveEncryptionPolicy(Protocol):
+    def resolve_campaign_archive_encryption(
+        self,
+        session: object | None = None,
+        *,
+        request: CampaignArchiveEncryptionRequest,
+    ) -> CampaignArchiveEncryptionDecision: ...
+
+
+def campaign_archive_encryption_policy(
+    registry: object | None,
+) -> CampaignArchiveEncryptionPolicy | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(CAPABILITY_POLICY_CAMPAIGN_ARCHIVE_ENCRYPTION)
+    ):
+        return None
+    capability = registry.capability(CAPABILITY_POLICY_CAMPAIGN_ARCHIVE_ENCRYPTION)
+    return (
+        capability
+        if isinstance(capability, CampaignArchiveEncryptionPolicy)
+        else None
+    )
+
+
 @dataclass(frozen=True, slots=True)
 class FunctionAssignmentGovernanceRequest:
     tenant_id: str
@@ -196,6 +421,20 @@ class FunctionAssignmentGovernanceRequest:
     context: Mapping[str, Any] = field(default_factory=dict)
 
 
+@dataclass(frozen=True, slots=True)
+class FunctionAssignmentEscalationRule:
+    step: FunctionAssignmentReviewStep
+    target_function_id: str
+    timeout_hours: int
+
+    def to_dict(self) -> dict[str, Any]:
+        return {
+            "step": self.step,
+            "target_function_id": self.target_function_id,
+            "timeout_hours": self.timeout_hours,
+        }
+
+
 @dataclass(frozen=True, slots=True)
 class FunctionAssignmentGovernanceDecision:
     allowed: bool
@@ -208,6 +447,10 @@ class FunctionAssignmentGovernanceDecision:
     separation_of_duties: bool = True
     quorum: int = 1
     maximum_validity_days: int | None = None
+    delegation_allowed: bool = False
+    maximum_delegation_depth: int = 0
+    maximum_delegated_validity_days: int | None = None
+    escalation_rules: tuple[FunctionAssignmentEscalationRule, ...] = ()
     request_expiry_hours: int = 336
     source_path: tuple[PolicySourceStep, ...] = ()
     requirements: tuple[str, ...] = ()
@@ -225,12 +468,24 @@ class FunctionAssignmentGovernanceDecision:
             "separation_of_duties": self.separation_of_duties,
             "quorum": self.quorum,
             "maximum_validity_days": self.maximum_validity_days,
+            "delegation_allowed": self.delegation_allowed,
+            "maximum_delegation_depth": self.maximum_delegation_depth,
+            "maximum_delegated_validity_days": (
+                self.maximum_delegated_validity_days
+            ),
+            "escalation_rules": [rule.to_dict() for rule in self.escalation_rules],
             "request_expiry_hours": self.request_expiry_hours,
             "source_path": [step.to_dict() for step in self.source_path],
             "requirements": list(self.requirements),
             "details": dict(self.details),
         }
 
+    def escalation_rule(
+        self,
+        step: FunctionAssignmentReviewStep,
+    ) -> FunctionAssignmentEscalationRule | None:
+        return next((rule for rule in self.escalation_rules if rule.step == step), None)
+
 
 @runtime_checkable
 class FunctionAssignmentGovernancePolicy(Protocol):

src/govoplan_core/core/postbox.py

@@ -15,6 +15,7 @@ CAPABILITY_POSTBOX_MESSAGES = "postbox.messages"
 CAPABILITY_POSTBOX_DELIVERY = "postbox.delivery"
 CAPABILITY_POSTBOX_EVIDENCE = "postbox.evidence"
 CAPABILITY_POSTBOX_ROUTING = "postbox.routing"
+CAPABILITY_POSTBOX_PORTAL = "postbox.portal_projection"
 
 PostboxAction = Literal[
     "discover",
@@ -171,6 +172,11 @@ class PostboxDirectoryEntryRef:
     template_revision_id: str | None = None
     holder_count: int = 0
     vacant: bool = True
+    encryption_profile: str = "plaintext_v1"
+    key_epoch: int = 1
+    encryption_vault_id: str | None = None
+    protection_policy: Mapping[str, object] = field(default_factory=dict)
+    grouping_policy: Mapping[str, object] = field(default_factory=dict)
     access: PostboxAccessDecisionRef | None = None
     resource_revision: int = 1
     etag: str | None = None
@@ -259,6 +265,9 @@ class PostboxMessageAuthoringRequest:
     idempotency_key: str
     subject: str
     body_text: str | None = None
+    ciphertext_ref: str | None = None
+    signed_manifest_ref: str | None = None
+    wrapped_keys: tuple[PostboxWrappedKeyRef, ...] = ()
     classification: str = "internal"
     participants: tuple[PostboxParticipantRef, ...] = ()
     attachments: tuple[PostboxAttachmentRef, ...] = ()
@@ -300,6 +309,7 @@ class PostboxDeliveryRequest:
     body_text: str | None = None
     sender_label: str | None = None
     classification: str = "internal"
+    action_required: bool = False
     participants: tuple[PostboxParticipantRef, ...] = ()
     attachments: tuple[PostboxAttachmentRef, ...] = ()
     expires_at: datetime | None = None
@@ -323,6 +333,14 @@ class PostboxDeliveryResult:
     evidence: Mapping[str, object] = field(default_factory=dict)
 
 
+@dataclass(frozen=True, slots=True)
+class PostboxPortalEntryRef:
+    postbox: PostboxDirectoryEntryRef
+    unread_count: int = 0
+    latest_message_at: datetime | None = None
+    route_path: str = "/postbox"
+
+
 class PostboxDeliveryRejected(RuntimeError):
     """A delivery was rejected before the provider accepted any effect."""
 
@@ -494,6 +512,31 @@ class PostboxRoutingProvider(Protocol):
     ) -> Mapping[str, object]:
         ...
 
+    def reconcile_notification_lifecycle(
+        self,
+        session: object,
+        *,
+        tenant_id: str | None = None,
+        limit: int = 50,
+    ) -> Mapping[str, object]:
+        """Reconcile assignment-derived Postbox notification facts."""
+
+        ...
+
+
+@runtime_checkable
+class PostboxPortalProjectionProvider(Protocol):
+    """Project portal-enabled Postboxes without transferring access ownership."""
+
+    def list_portal_entries(
+        self,
+        session: object,
+        principal: object,
+        *,
+        tenant_id: str,
+        limit: int = 100,
+    ) -> Sequence[PostboxPortalEntryRef]: ...
+
 
 def _postbox_provider(
     registry: object | None,
@@ -573,3 +616,18 @@ def postbox_routing_provider(
         provider_type=PostboxRoutingProvider,
     )
     return provider if isinstance(provider, PostboxRoutingProvider) else None
+
+
+def postbox_portal_projection_provider(
+    registry: object | None,
+) -> PostboxPortalProjectionProvider | None:
+    provider = _postbox_provider(
+        registry,
+        capability_name=CAPABILITY_POSTBOX_PORTAL,
+        provider_type=PostboxPortalProjectionProvider,
+    )
+    return (
+        provider
+        if isinstance(provider, PostboxPortalProjectionProvider)
+        else None
+    )

src/govoplan_core/core/principal_helpers.py

@@ -0,0 +1,37 @@
+"""Pure principal attribution mechanics, not authorization or tenant resolution.
+
+The two existing contracts intentionally differ in precedence and whitespace.
+Callers retain their own service-account, scope and resource-access decisions.
+"""
+
+
+def principal_actor_ids(principal: object) -> tuple[str, ...]:
+    """Account-first legacy IDs, unique in encounter order; retain nonblank text."""
+    user = getattr(principal, "user", None)
+    return tuple(
+        dict.fromkeys(
+            str(value)
+            for value in (
+                getattr(principal, "account_id", None),
+                getattr(principal, "identity_id", None),
+                getattr(principal, "membership_id", None),
+                getattr(user, "id", None),
+            )
+            if str(value or "").strip()
+        )
+    )
+
+
+def principal_user_first_actor(principal: object) -> str | None:
+    """First nonblank user/account/identity/membership ID, with trimmed text."""
+    user = getattr(principal, "user", None)
+    for value in (
+        getattr(user, "id", None),
+        getattr(principal, "account_id", None),
+        getattr(principal, "identity_id", None),
+        getattr(principal, "membership_id", None),
+    ):
+        candidate = str(value or "").strip()
+        if candidate:
+            return candidate
+    return None

src/govoplan_core/core/records.py

@@ -0,0 +1,369 @@
+from __future__ import annotations
+
+from collections.abc import Mapping, Sequence
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Literal, Protocol, runtime_checkable
+
+
+CAPABILITY_RECORDS_FILING = "records.filing"
+CAPABILITY_RECORD_SOURCE_PREFIX = "records.source."
+CAPABILITY_RECORD_ARCHIVE_PREFIX = "records.archive."
+
+RecordSourceAuthority = Literal[
+    "native_authoritative",
+    "external_authoritative",
+    "external_mirror",
+    "governed_sync",
+    "governance_overlay",
+    "linked_reference",
+]
+RecordArchiveOutcome = Literal["accepted", "rejected", "outcome_unknown"]
+
+_RECORD_SOURCE_AUTHORITIES = {
+    "native_authoritative",
+    "external_authoritative",
+    "external_mirror",
+    "governed_sync",
+    "governance_overlay",
+    "linked_reference",
+}
+_RECORD_ARCHIVE_OUTCOMES = {"accepted", "rejected", "outcome_unknown"}
+
+
+class RecordContractError(ValueError):
+    """Stable error for provider-neutral record filing operations."""
+
+
+@dataclass(frozen=True, slots=True)
+class RecordSourceLocator:
+    """Exact source revision requested for filing into a record."""
+
+    tenant_id: str
+    source_module: str
+    resource_type: str
+    resource_id: str
+    source_revision: str
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text_fields(
+            self,
+            "tenant_id",
+            "source_module",
+            "resource_type",
+            "resource_id",
+            "source_revision",
+        )
+        if len(self.resource_id) > 500 or len(self.source_revision) > 255:
+            raise RecordContractError("Record source identity is too long.")
+
+
+@dataclass(frozen=True, slots=True)
+class RecordSourceReference:
+    """Provider-resolved immutable source metadata safe to preserve in Records."""
+
+    locator: RecordSourceLocator
+    label: str
+    authority_mode: RecordSourceAuthority = "linked_reference"
+    content_sha256: str | None = None
+    content_type: str | None = None
+    size_bytes: int | None = None
+    valid_from: datetime | None = None
+    valid_to: datetime | None = None
+    recorded_at: datetime | None = None
+    launch_url: str | None = None
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        if not self.label.strip():
+            raise RecordContractError("Record source references require a label.")
+        if len(self.label) > 500:
+            raise RecordContractError(
+                "Record source labels are limited to 500 characters."
+            )
+        if self.size_bytes is not None and self.size_bytes < 0:
+            raise RecordContractError("Record source sizes cannot be negative.")
+        if self.content_sha256 is not None:
+            digest = self.content_sha256.removeprefix("sha256:")
+            if len(digest) != 64 or any(
+                character not in "0123456789abcdefABCDEF" for character in digest
+            ):
+                raise RecordContractError(
+                    "Record source SHA-256 digests must be hexadecimal."
+                )
+        if self.valid_from and self.valid_to and self.valid_to <= self.valid_from:
+            raise RecordContractError(
+                "Record source valid_to must be after valid_from."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class RecordFilingRequest:
+    tenant_id: str
+    record_id: str
+    source: RecordSourceLocator
+    purpose: str
+    filing_reason: str
+    idempotency_key: str
+    volume_id: str | None = None
+    relationship: str = "contains"
+    institutional_context: Mapping[str, object] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text_fields(
+            self,
+            "tenant_id",
+            "record_id",
+            "purpose",
+            "filing_reason",
+            "idempotency_key",
+            "relationship",
+        )
+        if self.source.tenant_id != self.tenant_id:
+            raise RecordContractError("Record filing cannot cross tenants.")
+        if len(self.purpose) > 255 or len(self.filing_reason) > 2_000:
+            raise RecordContractError("Record filing purpose or reason is too long.")
+        if len(self.idempotency_key) > 255:
+            raise RecordContractError(
+                "Record filing idempotency keys are limited to 255 characters."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class RecordFilingResult:
+    record_id: str
+    item_id: str
+    sequence: int
+    source: RecordSourceReference
+    filed_at: datetime
+    replayed: bool = False
+
+
+@dataclass(frozen=True, slots=True)
+class RecordTransferPackage:
+    """Exact, digest-bound package prepared by Records for one provider profile."""
+
+    tenant_id: str
+    package_id: str
+    record_id: str
+    record_revision: int
+    profile: str
+    manifest_sha256: str
+    manifest: Mapping[str, object]
+
+    def __post_init__(self) -> None:
+        _require_text_fields(
+            self,
+            "tenant_id",
+            "package_id",
+            "record_id",
+            "profile",
+            "manifest_sha256",
+        )
+        if self.record_revision < 1:
+            raise RecordContractError("Record transfer revisions must be positive.")
+        _require_sha256(self.manifest_sha256, "Record transfer manifest")
+
+
+@dataclass(frozen=True, slots=True)
+class RecordArchiveProviderState:
+    provider_id: str
+    label: str
+    profiles: tuple[str, ...]
+    authority_modes: tuple[RecordSourceAuthority, ...]
+    healthy: bool
+    checked_at: datetime
+    last_success_at: datetime | None = None
+    freshness_seconds: int | None = None
+    limitations: tuple[str, ...] = ()
+    simulated: bool = False
+
+    def __post_init__(self) -> None:
+        _require_text_fields(self, "provider_id", "label")
+        if not self.profiles or any(not item.strip() for item in self.profiles):
+            raise RecordContractError(
+                "Record archive providers require at least one profile."
+            )
+        if not self.authority_modes:
+            raise RecordContractError(
+                "Record archive providers require an authority mode."
+            )
+        if any(mode not in _RECORD_SOURCE_AUTHORITIES for mode in self.authority_modes):
+            raise RecordContractError(
+                "Record archive providers declared an invalid authority mode."
+            )
+        if self.freshness_seconds is not None and self.freshness_seconds < 0:
+            raise RecordContractError(
+                "Record archive provider freshness cannot be negative."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class RecordArchiveTransferRequest:
+    package: RecordTransferPackage
+    purpose: str
+    idempotency_key: str
+    institutional_context: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text_fields(self, "purpose", "idempotency_key")
+        if len(self.purpose) > 255 or len(self.idempotency_key) > 255:
+            raise RecordContractError(
+                "Record archive purpose or idempotency key is too long."
+            )
+
+
+@dataclass(frozen=True, slots=True)
+class RecordArchiveReceipt:
+    provider_id: str
+    package_id: str
+    outcome: RecordArchiveOutcome
+    observed_at: datetime
+    receipt_sha256: str
+    external_reference: str | None = None
+    retry_safe: bool = False
+    simulated: bool = False
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _require_text_fields(
+            self,
+            "provider_id",
+            "package_id",
+            "outcome",
+            "receipt_sha256",
+        )
+        _require_sha256(self.receipt_sha256, "Record archive receipt")
+        if self.outcome not in _RECORD_ARCHIVE_OUTCOMES:
+            raise RecordContractError("Record archive receipt outcome is invalid.")
+        if self.outcome == "outcome_unknown" and self.retry_safe:
+            raise RecordContractError(
+                "Unknown archive outcomes cannot be declared retry-safe."
+            )
+
+
+@runtime_checkable
+class RecordSourceProvider(Protocol):
+    provider_id: str
+
+    def resource_types(self) -> Sequence[str]: ...
+
+    def resolve(
+        self,
+        session: object,
+        principal: object,
+        *,
+        locator: RecordSourceLocator,
+        purpose: str,
+    ) -> RecordSourceReference:
+        """Resolve one currently authorized, exact source revision."""
+
+
+@runtime_checkable
+class RecordFilingService(Protocol):
+    def file(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: RecordFilingRequest,
+    ) -> RecordFilingResult: ...
+
+
+@runtime_checkable
+class RecordArchiveProvider(Protocol):
+    provider_id: str
+
+    def state(self) -> RecordArchiveProviderState: ...
+
+    def dispatch(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: RecordArchiveTransferRequest,
+    ) -> RecordArchiveReceipt:
+        """Dispatch one prepared package without retrying an unknown outcome."""
+
+
+def record_source_capability(source_module: str) -> str:
+    normalized = _capability_suffix(source_module, "source module")
+    return f"{CAPABILITY_RECORD_SOURCE_PREFIX}{normalized}"
+
+
+def record_archive_capability(provider_id: str) -> str:
+    normalized = _capability_suffix(provider_id, "archive provider")
+    return f"{CAPABILITY_RECORD_ARCHIVE_PREFIX}{normalized}"
+
+
+def record_source_capabilities(registry: object | None) -> tuple[str, ...]:
+    if registry is None or not hasattr(registry, "capability_names"):
+        return ()
+    return tuple(
+        name
+        for name in registry.capability_names()
+        if str(name).startswith(CAPABILITY_RECORD_SOURCE_PREFIX)
+    )
+
+
+def record_archive_capabilities(registry: object | None) -> tuple[str, ...]:
+    if registry is None or not hasattr(registry, "capability_names"):
+        return ()
+    return tuple(
+        name
+        for name in registry.capability_names()
+        if str(name).startswith(CAPABILITY_RECORD_ARCHIVE_PREFIX)
+    )
+
+
+def _require_text_fields(value: object, *field_names: str) -> None:
+    for field_name in field_names:
+        if not str(getattr(value, field_name, "") or "").strip():
+            raise RecordContractError(
+                f"Record contract field {field_name} is required."
+            )
+
+
+def _capability_suffix(value: str, label: str) -> str:
+    normalized = value.strip().lower()
+    if not normalized or any(
+        character not in "abcdefghijklmnopqrstuvwxyz0123456789_-"
+        for character in normalized
+    ):
+        raise RecordContractError(f"Record {label} identifiers are invalid.")
+    return normalized
+
+
+def _require_sha256(value: str, label: str) -> None:
+    digest = value.removeprefix("sha256:")
+    if len(digest) != 64 or any(
+        character not in "0123456789abcdefABCDEF" for character in digest
+    ):
+        raise RecordContractError(f"{label} SHA-256 must be hexadecimal.")
+
+
+__all__ = [
+    "CAPABILITY_RECORD_ARCHIVE_PREFIX",
+    "CAPABILITY_RECORDS_FILING",
+    "CAPABILITY_RECORD_SOURCE_PREFIX",
+    "RecordArchiveOutcome",
+    "RecordArchiveProvider",
+    "RecordArchiveProviderState",
+    "RecordArchiveReceipt",
+    "RecordArchiveTransferRequest",
+    "RecordContractError",
+    "RecordFilingRequest",
+    "RecordFilingResult",
+    "RecordFilingService",
+    "RecordSourceAuthority",
+    "RecordSourceLocator",
+    "RecordSourceProvider",
+    "RecordSourceReference",
+    "RecordTransferPackage",
+    "record_archive_capabilities",
+    "record_archive_capability",
+    "record_source_capabilities",
+    "record_source_capability",
+]

src/govoplan_core/core/runtime_work.py

@@ -0,0 +1,258 @@
+from __future__ import annotations
+
+from collections.abc import Callable, Mapping
+from datetime import datetime
+from typing import Any
+
+from govoplan_core.core.operations import (
+    RuntimeWorkState,
+    RuntimeWorkStatus,
+    RuntimeWorkStatusContext,
+)
+from govoplan_core.settings import settings
+
+
+QueueDepthReader = Callable[[list[str]], Mapping[str, int | None]]
+
+
+def celery_runtime_work_status(context: RuntimeWorkStatusContext) -> RuntimeWorkStatus:
+    """Observe the built-in Celery backend without exposing it to Ops."""
+
+    queues = _configured_queues()
+    backend_configured = bool(str(settings.redis_url or "").strip())
+    if not settings.celery_enabled:
+        return _status(
+            context,
+            enabled=False,
+            configured=backend_configured,
+            state="disabled",
+            detail="Background workers are intentionally disabled.",
+            active_workers=0,
+            queue_depths={queue: None for queue in queues},
+            active_work=0,
+            reserved_work=0,
+            guidance=(
+                "Synchronous development paths may be used in development. "
+                "Enable and monitor workers before production queue-backed work."
+                if context.profile in {"development", "local-dev"}
+                else "Enable a worker backend before accepting queue-backed work."
+            ),
+        )
+    if not backend_configured or not queues:
+        return _status(
+            context,
+            enabled=True,
+            configured=False,
+            state="unconfigured",
+            detail="Background workers are enabled but their backend or queue list is not configured.",
+            queue_depths={queue: None for queue in queues},
+            guidance="Configure the broker and an explicit queue list, then start the required worker pools.",
+        )
+
+    try:
+        from govoplan_core.celery_app import celery
+
+        inspector = celery.control.inspect(timeout=0.75)
+        return collect_celery_runtime_work_status(
+            context,
+            inspector=inspector,
+            queues=queues,
+            queue_depth_reader=_redis_queue_depths,
+        )
+    except Exception:  # noqa: BLE001 - status must isolate and sanitize provider failures.
+        return _status(
+            context,
+            enabled=True,
+            configured=True,
+            state="unreachable",
+            detail="The configured worker backend did not return bounded status evidence.",
+            queue_depths={queue: None for queue in queues},
+            guidance="Verify broker reachability and worker processes; do not infer health from missing metrics.",
+        )
+
+
+def collect_celery_runtime_work_status(
+    context: RuntimeWorkStatusContext,
+    *,
+    inspector: Any,
+    queues: list[str],
+    queue_depth_reader: QueueDepthReader,
+) -> RuntimeWorkStatus:
+    replies = inspector.ping() or {}
+    if not isinstance(replies, Mapping) or not replies:
+        state = "starting" if _fresh_worker_nodes(context) else "unreachable"
+        return _status(
+            context,
+            enabled=True,
+            configured=True,
+            state=state,
+            detail=(
+                "Worker processes are starting but have not answered the bounded status probe."
+                if state == "starting"
+                else "No configured worker answered the bounded status probe."
+            ),
+            active_workers=0,
+            queue_depths={queue: None for queue in queues},
+            guidance="Wait for startup or verify worker and broker connectivity.",
+        )
+
+    active_queues_by_worker = inspector.active_queues() or {}
+    active_by_worker = inspector.active() or {}
+    reserved_by_worker = inspector.reserved() or {}
+    active_queues = sorted(
+        {
+            str(queue.get("name"))
+            for worker_queues in active_queues_by_worker.values()
+            if isinstance(worker_queues, list)
+            for queue in worker_queues
+            if isinstance(queue, Mapping) and queue.get("name")
+        }
+    )
+    missing_queues = sorted(set(queues) - set(active_queues))
+    active_work = _task_count(active_by_worker)
+    reserved_work = _task_count(reserved_by_worker)
+    try:
+        measured_depths = dict(queue_depth_reader(queues))
+    except Exception:  # noqa: BLE001 - queue depth remains explicitly unsupported.
+        measured_depths = {}
+    queue_depths = {
+        queue: _bounded_count(measured_depths.get(queue)) for queue in queues
+    }
+    known_depth = sum(value for value in queue_depths.values() if value is not None)
+    worker_nodes = _worker_nodes(context)
+    stale_nodes = [node for node in worker_nodes if node.get("stale") is True]
+    latest_heartbeat = _latest_heartbeat(worker_nodes)
+
+    if stale_nodes and len(stale_nodes) >= len(worker_nodes) > 0:
+        state = "stale"
+        detail = "All registered worker heartbeats are stale."
+        guidance = "Restore worker heartbeats or replace the stale worker incarnations."
+    elif missing_queues or stale_nodes:
+        state = "degraded"
+        detail = "Worker status is partial: a queue lacks a consumer or a registered worker is stale."
+        guidance = "Restore the missing queue consumers and investigate stale worker heartbeats."
+    elif active_work + reserved_work + known_depth > 0:
+        state = "busy"
+        detail = "Workers are processing or waiting to process queued work."
+        guidance = "Monitor queue age and failures; scale only within configured provider limits."
+    elif all(value is not None for value in queue_depths.values()):
+        state = "idle"
+        detail = "Workers are available and all measured queues are empty."
+        guidance = "No action is required."
+    else:
+        state = "healthy"
+        detail = "Workers answered and cover all configured queues; queue depth is unavailable."
+        guidance = "Treat queue depth as unavailable, not empty."
+
+    return _status(
+        context,
+        enabled=True,
+        configured=True,
+        state=state,
+        detail=detail,
+        active_workers=len(replies),
+        last_heartbeat_at=latest_heartbeat,
+        queue_depths=queue_depths,
+        active_work=active_work,
+        reserved_work=reserved_work,
+        guidance=guidance,
+    )
+
+
+def _status(
+    context: RuntimeWorkStatusContext,
+    *,
+    enabled: bool,
+    configured: bool,
+    state: RuntimeWorkState,
+    detail: str,
+    active_workers: int | None = None,
+    last_heartbeat_at: datetime | None = None,
+    queue_depths: Mapping[str, int | None] | None = None,
+    active_work: int | None = None,
+    reserved_work: int | None = None,
+    guidance: str,
+) -> RuntimeWorkStatus:
+    return RuntimeWorkStatus(
+        provider_id="core.celery",
+        label="Background workers",
+        backend="Celery",
+        enabled=enabled,
+        configured=configured,
+        state=state,
+        detail=detail,
+        observed_at=context.observed_at,
+        active_workers=active_workers,
+        last_heartbeat_at=last_heartbeat_at,
+        queue_depths=queue_depths or {},
+        active_work=active_work,
+        reserved_work=reserved_work,
+        failures=None,
+        stale_after_seconds=context.stale_after_seconds,
+        guidance=guidance,
+    )
+
+
+def _configured_queues() -> list[str]:
+    return sorted(
+        {
+            item.strip()
+            for item in str(settings.celery_queues or "").split(",")
+            if item.strip()
+        }
+    )
+
+
+def _redis_queue_depths(queues: list[str]) -> Mapping[str, int | None]:
+    from redis import Redis
+
+    client = Redis.from_url(
+        settings.redis_url,
+        socket_connect_timeout=0.75,
+        socket_timeout=0.75,
+    )
+    pipeline = client.pipeline(transaction=False)
+    for queue in queues:
+        pipeline.llen(queue)
+    values = pipeline.execute()
+    return {
+        queue: _bounded_count(value)
+        for queue, value in zip(queues, values, strict=True)
+    }
+
+
+def _task_count(tasks_by_worker: object) -> int:
+    if not isinstance(tasks_by_worker, Mapping):
+        return 0
+    return sum(
+        len(tasks) for tasks in tasks_by_worker.values() if isinstance(tasks, list)
+    )
+
+
+def _bounded_count(value: object) -> int | None:
+    if isinstance(value, bool) or not isinstance(value, int | float):
+        return None
+    return max(0, int(value))
+
+
+def _worker_nodes(context: RuntimeWorkStatusContext) -> list[Mapping[str, object]]:
+    return [node for node in context.runtime_nodes if node.get("role") == "worker"]
+
+
+def _fresh_worker_nodes(context: RuntimeWorkStatusContext) -> list[Mapping[str, object]]:
+    return [node for node in _worker_nodes(context) if node.get("stale") is not True]
+
+
+def _latest_heartbeat(nodes: list[Mapping[str, object]]) -> datetime | None:
+    values: list[datetime] = []
+    for node in nodes:
+        raw = node.get("last_heartbeat_at")
+        if isinstance(raw, datetime):
+            values.append(raw)
+            continue
+        if isinstance(raw, str):
+            try:
+                values.append(datetime.fromisoformat(raw.replace("Z", "+00:00")))
+            except ValueError:
+                continue
+    return max(values) if values else None

src/govoplan_core/core/semantic_documentation.py

@@ -0,0 +1,621 @@
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+from collections.abc import Mapping, Sequence
+from dataclasses import dataclass, field
+from typing import Literal, Protocol, runtime_checkable
+
+
+SEMANTIC_DOCUMENTATION_SUBJECT_CAPABILITY_PREFIX = (
+    "documentation.semantic_subjects."
+)
+SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION = "1"
+
+SemanticDocumentationSubjectAvailability = Literal[
+    "available",
+    "changed",
+    "superseded",
+    "missing",
+    "temporarily_unavailable",
+]
+
+_MODULE_ID_RE = re.compile(r"^[a-z][a-z0-9_]{0,79}$")
+_KIND_RE = re.compile(r"^[a-z][a-z0-9_.-]{0,119}$")
+_IDENTIFIER_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9_.:@-]{0,254}$")
+_LOCALE_RE = re.compile(r"^[A-Za-z]{2,3}(?:-[A-Za-z0-9]{2,8})*$")
+_REASON_CODE_RE = re.compile(r"^[a-z][a-z0-9_]{0,79}$")
+_SHA256_RE = re.compile(r"^(?:sha256:)?[0-9a-fA-F]{64}$")
+
+
+class SemanticDocumentationContractError(ValueError):
+    """Raised when a semantic-documentation subject violates the Core contract."""
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectAnchor:
+    kind: str
+    id: str
+
+    def __post_init__(self) -> None:
+        _require_match(self.kind, _KIND_RE, "Semantic subject anchor kind")
+        _require_match(self.id, _IDENTIFIER_RE, "Semantic subject anchor id")
+
+    def to_dict(self) -> dict[str, str]:
+        return {"kind": self.kind, "id": self.id}
+
+    @classmethod
+    def from_mapping(
+        cls, value: Mapping[str, object]
+    ) -> SemanticDocumentationSubjectAnchor:
+        _require_keys(value, {"kind", "id"}, "Semantic subject anchor")
+        return cls(kind=_required_text(value, "kind"), id=_required_text(value, "id"))
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectReference:
+    module_id: str
+    tenant_id: str
+    subject_kind: str
+    subject_id: str
+    anchor: SemanticDocumentationSubjectAnchor | None = None
+    observed_revision: str | None = None
+    observed_fingerprint: str | None = None
+
+    def __post_init__(self) -> None:
+        _require_match(self.module_id, _MODULE_ID_RE, "Semantic subject module id")
+        _require_match(self.tenant_id, _IDENTIFIER_RE, "Semantic subject tenant id")
+        _require_match(self.subject_kind, _KIND_RE, "Semantic subject kind")
+        _require_match(self.subject_id, _IDENTIFIER_RE, "Semantic subject id")
+        _optional_text(self.observed_revision, "Semantic subject observed revision", 255)
+        if self.observed_fingerprint is not None and not _SHA256_RE.fullmatch(
+            self.observed_fingerprint
+        ):
+            raise SemanticDocumentationContractError(
+                "Semantic subject observed fingerprint must be a SHA-256 digest."
+            )
+
+    @property
+    def stable_key(self) -> str:
+        identity = {
+            "anchor": self.anchor.to_dict() if self.anchor else None,
+            "module_id": self.module_id,
+            "subject_id": self.subject_id,
+            "subject_kind": self.subject_kind,
+            "tenant_id": self.tenant_id,
+        }
+        encoded = json.dumps(
+            identity, ensure_ascii=False, sort_keys=True, separators=(",", ":")
+        ).encode("utf-8")
+        return f"sha256:{hashlib.sha256(encoded).hexdigest()}"
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "module_id": self.module_id,
+            "tenant_id": self.tenant_id,
+            "subject_kind": self.subject_kind,
+            "subject_id": self.subject_id,
+            "anchor": self.anchor.to_dict() if self.anchor else None,
+            "observed_revision": self.observed_revision,
+            "observed_fingerprint": self.observed_fingerprint,
+        }
+
+    @classmethod
+    def from_mapping(
+        cls, value: Mapping[str, object]
+    ) -> SemanticDocumentationSubjectReference:
+        _require_keys(
+            value,
+            {
+                "module_id",
+                "tenant_id",
+                "subject_kind",
+                "subject_id",
+                "anchor",
+                "observed_revision",
+                "observed_fingerprint",
+            },
+            "Semantic subject reference",
+        )
+        raw_anchor = value.get("anchor")
+        if raw_anchor is not None and not isinstance(raw_anchor, Mapping):
+            raise SemanticDocumentationContractError(
+                "Semantic subject anchor must be an object."
+            )
+        return cls(
+            module_id=_required_text(value, "module_id"),
+            tenant_id=_required_text(value, "tenant_id"),
+            subject_kind=_required_text(value, "subject_kind"),
+            subject_id=_required_text(value, "subject_id"),
+            anchor=(
+                SemanticDocumentationSubjectAnchor.from_mapping(raw_anchor)
+                if isinstance(raw_anchor, Mapping)
+                else None
+            ),
+            observed_revision=_mapping_optional_text(value, "observed_revision"),
+            observed_fingerprint=_mapping_optional_text(
+                value, "observed_fingerprint"
+            ),
+        )
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationBreadcrumb:
+    label: str
+    subject_kind: str
+    subject_id: str
+    anchor: SemanticDocumentationSubjectAnchor | None = None
+
+    def __post_init__(self) -> None:
+        _required_bounded_text(self.label, "Semantic subject breadcrumb label", 300)
+        _require_match(self.subject_kind, _KIND_RE, "Semantic breadcrumb kind")
+        _require_match(self.subject_id, _IDENTIFIER_RE, "Semantic breadcrumb id")
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "label": self.label,
+            "subject_kind": self.subject_kind,
+            "subject_id": self.subject_id,
+            "anchor": self.anchor.to_dict() if self.anchor else None,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectDescriptor:
+    reference: SemanticDocumentationSubjectReference
+    labels: Mapping[str, str]
+    descriptions: Mapping[str, str] = field(default_factory=dict)
+    breadcrumbs: tuple[SemanticDocumentationBreadcrumb, ...] = ()
+    route: str | None = None
+    route_anchor: str | None = None
+    audience: tuple[str, ...] = ()
+    classification: str = "internal"
+    required_scopes: tuple[str, ...] = ()
+
+    def __post_init__(self) -> None:
+        if not self.reference.observed_revision:
+            raise SemanticDocumentationContractError(
+                "Semantic subject descriptors require a current revision."
+            )
+        if not self.reference.observed_fingerprint:
+            raise SemanticDocumentationContractError(
+                "Semantic subject descriptors require a current fingerprint."
+            )
+        _localized_text(self.labels, "Semantic subject labels", required=True, limit=300)
+        _localized_text(
+            self.descriptions,
+            "Semantic subject descriptions",
+            required=False,
+            limit=2_000,
+        )
+        if len(self.breadcrumbs) > 32:
+            raise SemanticDocumentationContractError(
+                "Semantic subject breadcrumbs are limited to 32 items."
+            )
+        if self.route is not None:
+            _optional_text(self.route, "Semantic subject route", 2_000)
+            if not self.route.startswith("/") or self.route.startswith("//"):
+                raise SemanticDocumentationContractError(
+                    "Semantic subject routes must be local absolute paths."
+                )
+        if self.route_anchor is not None:
+            _require_match(
+                self.route_anchor,
+                _IDENTIFIER_RE,
+                "Semantic subject route anchor",
+            )
+        _text_tuple(self.audience, "Semantic subject audience", maximum=32)
+        _required_bounded_text(
+            self.classification, "Semantic subject classification", 120
+        )
+        _text_tuple(
+            self.required_scopes, "Semantic subject required scopes", maximum=64
+        )
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "reference": self.reference.to_dict(),
+            "labels": dict(self.labels),
+            "descriptions": dict(self.descriptions),
+            "breadcrumbs": [item.to_dict() for item in self.breadcrumbs],
+            "route": self.route,
+            "route_anchor": self.route_anchor,
+            "audience": list(self.audience),
+            "classification": self.classification,
+            "required_scopes": list(self.required_scopes),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectResolution:
+    requested_reference: SemanticDocumentationSubjectReference
+    availability: SemanticDocumentationSubjectAvailability
+    subject: SemanticDocumentationSubjectDescriptor | None = None
+    superseded_by: SemanticDocumentationSubjectReference | None = None
+    reason_code: str | None = None
+
+    def __post_init__(self) -> None:
+        if self.reason_code is not None:
+            _require_match(
+                self.reason_code, _REASON_CODE_RE, "Semantic resolution reason code"
+            )
+        if self.availability in {"available", "changed"}:
+            if self.subject is None:
+                raise SemanticDocumentationContractError(
+                    f"Semantic subject {self.availability} resolutions require a descriptor."
+                )
+            if (
+                self.subject.reference.stable_key
+                != self.requested_reference.stable_key
+            ):
+                raise SemanticDocumentationContractError(
+                    "Semantic subject resolution changed the requested identity."
+                )
+            changed = _reference_changed(
+                self.requested_reference, self.subject.reference
+            )
+            if self.availability == "available" and changed:
+                raise SemanticDocumentationContractError(
+                    "Changed semantic subjects must use the changed availability."
+                )
+            if self.availability == "changed" and not changed:
+                raise SemanticDocumentationContractError(
+                    "Changed semantic subject resolutions require a revision or fingerprint change."
+                )
+        elif self.subject is not None:
+            raise SemanticDocumentationContractError(
+                f"Semantic subject {self.availability} resolutions cannot include a descriptor."
+            )
+        if self.availability == "superseded":
+            if self.superseded_by is None:
+                raise SemanticDocumentationContractError(
+                    "Superseded semantic subjects require a replacement reference."
+                )
+        elif self.superseded_by is not None:
+            raise SemanticDocumentationContractError(
+                "Only superseded semantic subjects may declare a replacement."
+            )
+        if self.availability in {"missing", "temporarily_unavailable"} and not self.reason_code:
+            raise SemanticDocumentationContractError(
+                f"Semantic subject {self.availability} resolutions require a reason code."
+            )
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "requested_reference": self.requested_reference.to_dict(),
+            "availability": self.availability,
+            "subject": self.subject.to_dict() if self.subject else None,
+            "superseded_by": (
+                self.superseded_by.to_dict() if self.superseded_by else None
+            ),
+            "reason_code": self.reason_code,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectQuery:
+    tenant_id: str
+    query: str = ""
+    subject_kinds: tuple[str, ...] = ()
+    limit: int = 50
+    cursor: str | None = None
+
+    def __post_init__(self) -> None:
+        _require_match(self.tenant_id, _IDENTIFIER_RE, "Semantic query tenant id")
+        if not isinstance(self.query, str) or len(self.query) > 300:
+            raise SemanticDocumentationContractError(
+                "Semantic subject query must be text of at most 300 characters."
+            )
+        if self.query:
+            _required_bounded_text(self.query, "Semantic subject query", 300)
+        if not 1 <= self.limit <= 200:
+            raise SemanticDocumentationContractError(
+                "Semantic subject query limit must be between 1 and 200."
+            )
+        _text_tuple(self.subject_kinds, "Semantic query subject kinds", maximum=100)
+        for kind in self.subject_kinds:
+            _require_match(kind, _KIND_RE, "Semantic query subject kind")
+        _optional_text(self.cursor, "Semantic query cursor", 1_000)
+
+
+@dataclass(frozen=True, slots=True)
+class SemanticDocumentationSubjectPage:
+    subjects: tuple[SemanticDocumentationSubjectDescriptor, ...] = ()
+    next_cursor: str | None = None
+    has_more: bool = False
+
+    def __post_init__(self) -> None:
+        keys = tuple(item.reference.stable_key for item in self.subjects)
+        if len(keys) != len(set(keys)):
+            raise SemanticDocumentationContractError(
+                "Semantic subject pages cannot contain duplicate identities."
+            )
+        _optional_text(self.next_cursor, "Semantic subject page cursor", 1_000)
+        if self.has_more and not self.next_cursor:
+            raise SemanticDocumentationContractError(
+                "Semantic subject pages with more results require a cursor."
+            )
+
+
+@runtime_checkable
+class SemanticDocumentationSubjectProvider(Protocol):
+    provider_id: str
+    module_id: str
+    contract_version: str
+
+    def list_subjects(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: SemanticDocumentationSubjectQuery,
+    ) -> SemanticDocumentationSubjectPage: ...
+
+    def resolve_subject(
+        self,
+        session: object,
+        principal: object,
+        *,
+        reference: SemanticDocumentationSubjectReference,
+    ) -> SemanticDocumentationSubjectResolution | None:
+        """Return None when the principal may not know whether a subject exists."""
+
+
+def semantic_documentation_subject_capability(module_id: str) -> str:
+    _require_match(module_id, _MODULE_ID_RE, "Semantic subject module id")
+    return f"{SEMANTIC_DOCUMENTATION_SUBJECT_CAPABILITY_PREFIX}{module_id}"
+
+
+def semantic_documentation_subject_provider_names(
+    registry: object | None,
+) -> tuple[str, ...]:
+    if registry is None or not hasattr(registry, "capability_names"):
+        return ()
+    return tuple(
+        str(name)
+        for name in registry.capability_names()
+        if str(name).startswith(SEMANTIC_DOCUMENTATION_SUBJECT_CAPABILITY_PREFIX)
+    )
+
+
+def semantic_documentation_subject_providers(
+    registry: object | None,
+) -> tuple[tuple[str, SemanticDocumentationSubjectProvider], ...]:
+    if registry is None or not hasattr(registry, "capability"):
+        return ()
+    providers: list[tuple[str, SemanticDocumentationSubjectProvider]] = []
+    for capability_name in semantic_documentation_subject_provider_names(registry):
+        module_id = capability_name.removeprefix(
+            SEMANTIC_DOCUMENTATION_SUBJECT_CAPABILITY_PREFIX
+        )
+        provider = registry.capability(capability_name)
+        if not isinstance(provider, SemanticDocumentationSubjectProvider):
+            raise TypeError(
+                f"Invalid semantic-documentation provider capability: {capability_name}"
+            )
+        if provider.module_id != module_id:
+            raise SemanticDocumentationContractError(
+                f"Semantic provider module {provider.module_id!r} does not match "
+                f"capability {capability_name!r}."
+            )
+        if (
+            provider.contract_version
+            != SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION
+        ):
+            raise SemanticDocumentationContractError(
+                f"Unsupported semantic-documentation provider contract: "
+                f"{provider.contract_version!r}."
+            )
+        providers.append((module_id, provider))
+    return tuple(providers)
+
+
+def list_semantic_documentation_subjects(
+    registry: object | None,
+    session: object,
+    principal: object,
+    *,
+    request: SemanticDocumentationSubjectQuery,
+) -> tuple[tuple[str, SemanticDocumentationSubjectPage], ...]:
+    if _principal_tenant_id(principal) != request.tenant_id:
+        return ()
+    pages: list[tuple[str, SemanticDocumentationSubjectPage]] = []
+    for module_id, provider in semantic_documentation_subject_providers(registry):
+        page = provider.list_subjects(
+            session,
+            principal,
+            request=request,
+        )
+        if any(
+            subject.reference.module_id != module_id
+            or subject.reference.tenant_id != request.tenant_id
+            for subject in page.subjects
+        ):
+            raise SemanticDocumentationContractError(
+                f"Semantic provider {module_id!r} returned a foreign subject."
+            )
+        pages.append((module_id, page))
+    return tuple(pages)
+
+
+def resolve_semantic_documentation_subject(
+    registry: object | None,
+    session: object,
+    principal: object,
+    *,
+    reference: SemanticDocumentationSubjectReference,
+) -> SemanticDocumentationSubjectResolution | None:
+    if _principal_tenant_id(principal) != reference.tenant_id:
+        return None
+    capability_name = semantic_documentation_subject_capability(reference.module_id)
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not registry.has_capability(capability_name)
+    ):
+        return SemanticDocumentationSubjectResolution(
+            requested_reference=reference,
+            availability="temporarily_unavailable",
+            reason_code="provider_unavailable",
+        )
+    provider = registry.capability(capability_name)
+    if not isinstance(provider, SemanticDocumentationSubjectProvider):
+        raise TypeError(
+            f"Invalid semantic-documentation provider capability: {capability_name}"
+        )
+    if (
+        provider.module_id != reference.module_id
+        or provider.contract_version
+        != SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION
+    ):
+        raise SemanticDocumentationContractError(
+            f"Semantic provider {capability_name!r} does not match the Core contract."
+        )
+    result = provider.resolve_subject(
+        session,
+        principal,
+        reference=reference,
+    )
+    if result is None:
+        return None
+    if result.requested_reference != reference:
+        raise SemanticDocumentationContractError(
+            "Semantic provider returned a resolution for another reference."
+        )
+    return result
+
+
+def semantic_documentation_fingerprint(value: object) -> str:
+    try:
+        encoded = json.dumps(
+            value,
+            ensure_ascii=False,
+            sort_keys=True,
+            separators=(",", ":"),
+            allow_nan=False,
+        ).encode("utf-8")
+    except (TypeError, ValueError) as exc:
+        raise SemanticDocumentationContractError(
+            "Semantic fingerprint input must be canonical JSON data."
+        ) from exc
+    return f"sha256:{hashlib.sha256(encoded).hexdigest()}"
+
+
+def _reference_changed(
+    requested: SemanticDocumentationSubjectReference,
+    current: SemanticDocumentationSubjectReference,
+) -> bool:
+    comparisons = (
+        (requested.observed_revision, current.observed_revision),
+        (requested.observed_fingerprint, current.observed_fingerprint),
+    )
+    return any(expected is not None and expected != actual for expected, actual in comparisons)
+
+
+def _principal_tenant_id(principal: object) -> str:
+    return str(getattr(principal, "tenant_id", "") or "")
+
+
+def _localized_text(
+    values: Mapping[str, str],
+    label: str,
+    *,
+    required: bool,
+    limit: int,
+) -> None:
+    if required and not values:
+        raise SemanticDocumentationContractError(f"{label} are required.")
+    if len(values) > 20:
+        raise SemanticDocumentationContractError(f"{label} are limited to 20 locales.")
+    for locale, value in values.items():
+        if not _LOCALE_RE.fullmatch(str(locale)):
+            raise SemanticDocumentationContractError(
+                f"{label} contain an invalid locale: {locale!r}."
+            )
+        _required_bounded_text(value, f"{label} value", limit)
+
+
+def _text_tuple(values: Sequence[str], label: str, *, maximum: int) -> None:
+    if len(values) > maximum:
+        raise SemanticDocumentationContractError(
+            f"{label} are limited to {maximum} items."
+        )
+    normalized = tuple(str(value).strip() for value in values)
+    if any(not value or len(value) > 255 for value in normalized):
+        raise SemanticDocumentationContractError(
+            f"{label} must contain non-empty bounded text."
+        )
+    if len(normalized) != len(set(normalized)):
+        raise SemanticDocumentationContractError(f"{label} must be unique.")
+
+
+def _require_match(value: str, pattern: re.Pattern[str], label: str) -> None:
+    if not isinstance(value, str) or not pattern.fullmatch(value):
+        raise SemanticDocumentationContractError(f"{label} is invalid.")
+
+
+def _required_bounded_text(value: str, label: str, limit: int) -> None:
+    if not isinstance(value, str) or not value.strip() or len(value) > limit:
+        raise SemanticDocumentationContractError(
+            f"{label} must be non-empty and at most {limit} characters."
+        )
+    if any(ord(character) < 32 and character not in "\n\t" for character in value):
+        raise SemanticDocumentationContractError(f"{label} contains control characters.")
+
+
+def _optional_text(value: str | None, label: str, limit: int) -> None:
+    if value is not None:
+        _required_bounded_text(value, label, limit)
+
+
+def _require_keys(
+    value: Mapping[str, object], allowed: set[str], label: str
+) -> None:
+    unexpected = sorted(str(key) for key in value if str(key) not in allowed)
+    if unexpected:
+        raise SemanticDocumentationContractError(
+            f"{label} contains unsupported fields: {', '.join(unexpected)}."
+        )
+
+
+def _required_text(value: Mapping[str, object], key: str) -> str:
+    result = value.get(key)
+    if not isinstance(result, str) or not result.strip():
+        raise SemanticDocumentationContractError(
+            f"Semantic subject field {key} is required."
+        )
+    return result
+
+
+def _mapping_optional_text(value: Mapping[str, object], key: str) -> str | None:
+    result = value.get(key)
+    if result is None:
+        return None
+    if not isinstance(result, str):
+        raise SemanticDocumentationContractError(
+            f"Semantic subject field {key} must be text."
+        )
+    return result
+
+
+__all__ = [
+    "SEMANTIC_DOCUMENTATION_SUBJECT_CAPABILITY_PREFIX",
+    "SEMANTIC_DOCUMENTATION_SUBJECT_CONTRACT_VERSION",
+    "SemanticDocumentationBreadcrumb",
+    "SemanticDocumentationContractError",
+    "SemanticDocumentationSubjectAnchor",
+    "SemanticDocumentationSubjectAvailability",
+    "SemanticDocumentationSubjectDescriptor",
+    "SemanticDocumentationSubjectPage",
+    "SemanticDocumentationSubjectProvider",
+    "SemanticDocumentationSubjectQuery",
+    "SemanticDocumentationSubjectReference",
+    "SemanticDocumentationSubjectResolution",
+    "list_semantic_documentation_subjects",
+    "resolve_semantic_documentation_subject",
+    "semantic_documentation_fingerprint",
+    "semantic_documentation_subject_capability",
+    "semantic_documentation_subject_provider_names",
+    "semantic_documentation_subject_providers",
+]

src/govoplan_core/core/tabular_sources.py

@@ -1,11 +1,15 @@
 from __future__ import annotations
 
 import csv
+import hashlib
 import io
+import json
+import math
 import re
-from collections.abc import Mapping, Sequence
+from collections.abc import Callable, Mapping, Sequence
 from dataclasses import dataclass, field
 from datetime import datetime
+from decimal import Decimal
 from typing import Literal, Protocol, runtime_checkable
 
 
@@ -17,6 +21,77 @@ DEFAULT_PREVIEW_TIMEOUT_MS = 2_000
 TabularSourceMode = Literal["live", "cached", "file_backed", "static"]
 TabularHealthStatus = Literal["healthy", "warning", "error", "unknown"]
 TabularDiagnosticSeverity = Literal["info", "warning", "error"]
+CsvValueMode = Literal["legacy_typed", "text"]
+
+
+@dataclass(frozen=True, slots=True)
+class TabularCsvSource:
+    """Original upload text, retained only with an explicitly durable import.
+
+    This is never catalogue metadata or transient preview retention. Owners
+    enforce access, size limits, lifecycle and export authorization separately.
+    """
+
+    text: str
+    delimiter: str = ","
+    value_mode: CsvValueMode = "legacy_typed"
+    parser_profile: str = "core.csv.v1"
+
+
+def csv_source_payload(source: TabularCsvSource, *, max_bytes: int = 5_000_000) -> dict[str, object]:
+    encoded = _csv_utf8_bytes(source.text)
+    if len(encoded) > max_bytes:
+        raise TabularSourceValidationError(f"Original CSV input is limited to {max_bytes:,} UTF-8 bytes.")
+    if source.value_mode not in {"text", "legacy_typed"} or len(source.delimiter) != 1:
+        raise TabularSourceValidationError("Invalid CSV source parsing options.")
+    return {
+        "text": source.text,
+        "delimiter": source.delimiter,
+        "value_mode": source.value_mode,
+        "parser_profile": source.parser_profile,
+        "sha256": hashlib.sha256(encoded).hexdigest(),
+        "byte_count": len(encoded),
+    }
+
+
+def csv_source_summary(payload: Mapping[str, object]) -> dict[str, object]:
+    """Allowlist the small, non-content evidence safe for catalogue DTOs."""
+    result = {key: payload[key] for key in ("delimiter", "value_mode", "parser_profile", "sha256", "byte_count")}
+    if "governance_history" in payload:
+        result["governance_sha256"] = hashlib.sha256(json.dumps(payload["governance_history"], sort_keys=True, separators=(",", ":"), allow_nan=False).encode("utf-8")).hexdigest()
+    return result
+
+
+def verified_csv_source_text(payload: Mapping[str, object], *, expected_summary: Mapping[str, object] | None = None) -> str:
+    text = payload.get("text")
+    if not isinstance(text, str):
+        raise TabularSourceUnavailableError("Original CSV source text is unavailable.")
+    try:
+        encoded = text.encode("utf-8")
+    except UnicodeError as exc:
+        raise TabularSourceUnavailableError("Original CSV source encoding is invalid.") from exc
+    if len(encoded) != payload.get("byte_count") or hashlib.sha256(encoded).hexdigest() != payload.get("sha256"):
+        raise TabularSourceUnavailableError("Original CSV source integrity verification failed.")
+    if expected_summary is not None:
+        try:
+            actual = json.dumps(csv_source_summary(payload), sort_keys=True, separators=(",", ":"), allow_nan=False)
+            expected = json.dumps(expected_summary, sort_keys=True, separators=(",", ":"), allow_nan=False)
+        except (KeyError, TypeError, ValueError) as exc:
+            raise TabularSourceUnavailableError("Original CSV source evidence is invalid.") from exc
+        if actual != expected:
+            raise TabularSourceUnavailableError("Original CSV source no longer matches its recorded evidence.")
+    return text
+
+
+def csv_projection_matches(expected: Sequence[Mapping[str, object]], actual: Sequence[Mapping[str, object]]) -> bool:
+    """CSV cells are scalar: booleans, integers and floats are not interchangeable."""
+    return len(expected) == len(actual) and all(
+        left.keys() == right.keys() and all(
+            type(value) is type(right[name]) and value == right[name]
+            for name, value in left.items()
+        )
+        for left, right in zip(expected, actual, strict=True)
+    )
 
 
 class TabularSourceError(ValueError):
@@ -39,29 +114,44 @@ class TabularSourceUnavailableError(TabularSourceError):
     pass
 
 
+def _csv_utf8_bytes(text: str) -> bytes:
+    try:
+        return text.encode("utf-8")
+    except UnicodeError as exc:
+        raise TabularSourceValidationError("CSV input must be valid Unicode encodable as UTF-8.") from exc
+
+
 def parse_tabular_csv(
     csv_text: str,
     *,
     delimiter: str = ",",
     max_rows: int = 10_000,
+    max_bytes: int = 5_000_000,
+    value_mode: CsvValueMode = "legacy_typed",
 ) -> tuple[Mapping[str, object], ...]:
-    """Parse a bounded CSV document into JSON-compatible tabular rows."""
+    """Parse CSV with an explicit lexical-text or backward-compatible typed mode."""
 
     if len(delimiter) != 1:
         raise TabularSourceValidationError("CSV delimiter must be one character.")
+    if value_mode not in {"legacy_typed", "text"}:
+        raise TabularSourceValidationError("Unsupported CSV value mode.")
+    if len(_csv_utf8_bytes(csv_text)) > max_bytes:
+        raise TabularSourceValidationError(f"CSV input is limited to {max_bytes:,} UTF-8 bytes.")
     try:
-        reader = csv.DictReader(io.StringIO(csv_text), delimiter=delimiter)
+        reader = csv.DictReader(io.StringIO(csv_text), delimiter=delimiter, strict=value_mode == "text")
         original_headers, normalized_headers = _csv_headers(reader.fieldnames)
         rows: list[dict[str, object]] = []
         for row in reader:
             _validate_csv_row_shape(row)
-            if _csv_row_is_empty(row, original_headers):
+            if value_mode == "text" and (None in row or any(row.get(header) is None for header in original_headers)):
+                raise TabularSourceValidationError("CSV text rows must have exactly the number of values defined by the header.")
+            if value_mode == "legacy_typed" and _csv_row_is_empty(row, original_headers):
                 continue
             if len(rows) >= max_rows:
                 raise TabularSourceValidationError(
                     f"CSV snapshots are limited to {max_rows:,} rows."
                 )
-            rows.append(_csv_row(row, original_headers, normalized_headers))
+            rows.append(_csv_row(row, original_headers, normalized_headers, value_mode=value_mode))
         return tuple(rows)
     except csv.Error as exc:
         raise TabularSourceValidationError(f"CSV input could not be parsed: {exc}") from exc
@@ -106,9 +196,11 @@ def _csv_row(
     row: Mapping[str | None, str | list[str] | None],
     original_headers: Sequence[str],
     normalized_headers: Sequence[str],
+    *,
+    value_mode: CsvValueMode = "legacy_typed",
 ) -> dict[str, object]:
     return {
-        normalized: _csv_scalar(value if isinstance(value, str) else None)
+        normalized: (value if value_mode == "text" else _csv_scalar(value if isinstance(value, str) else None))
         for original, normalized in zip(
             original_headers,
             normalized_headers,
@@ -128,9 +220,15 @@ def _csv_scalar(value: str | None) -> object:
     if lowered in {"true", "false"}:
         return lowered == "true"
     if re.fullmatch(r"-?(?:0|[1-9][0-9]*)", text):
+        try:
             return int(text)
+        except ValueError as exc:
+            raise TabularSourceValidationError("CSV integer exceeds the conversion limit; use text mode to preserve it.") from exc
     if re.fullmatch(r"-?(?:0|[1-9][0-9]*)\.[0-9]+", text):
-        return float(text)
+        value = float(text)
+        if not math.isfinite(value):
+            raise TabularSourceValidationError("CSV numeric value exceeds the finite number range; use text mode to preserve it.")
+        return value
     return text
 
 
@@ -141,6 +239,48 @@ class TabularColumn:
     nullable: bool = True
 
 
+def tabular_type_name(value: object, *, casefold_unknown: bool = False) -> str:
+    if isinstance(value, bool):
+        return "boolean"
+    if isinstance(value, int):
+        return "integer"
+    if isinstance(value, (float, Decimal)):
+        return "number"
+    if isinstance(value, str):
+        return "string"
+    if isinstance(value, list):
+        return "array"
+    if isinstance(value, dict):
+        return "object"
+    name = type(value).__name__
+    return name.casefold() if casefold_unknown else name.lower()
+
+
+def infer_tabular_schema(
+    rows: Sequence[Mapping[str, object]],
+    *,
+    type_name: Callable[[object], str] = tabular_type_name,
+) -> tuple[TabularColumn, ...]:
+    """Infer first-seen columns in one pass without retaining column values.
+
+    The classifier is explicit so legacy providers can preserve their exact
+    type naming. Missing keys and explicit None both make a column nullable.
+    """
+    states: dict[str, tuple[str | None, int]] = {}
+    for row in rows:
+        for name, value in row.items():
+            kind, concrete = states.get(name, (None, 0))
+            if value is not None:
+                value_kind = type_name(value)
+                kind = value_kind if kind is None else kind if kind == value_kind else "mixed"
+                concrete += 1
+            states[name] = (kind, concrete)
+    return tuple(
+        TabularColumn(name=name, data_type=kind if kind is not None else "unknown", nullable=concrete != len(rows))
+        for name, (kind, concrete) in states.items()
+    )
+
+
 @dataclass(frozen=True, slots=True)
 class TabularPushdown:
     projections: bool = False
@@ -221,6 +361,7 @@ class TabularSnapshotInput:
     rows: tuple[Mapping[str, object], ...]
     description: str | None = None
     metadata: Mapping[str, object] = field(default_factory=dict)
+    csv_source: TabularCsvSource | None = None
 
 
 @runtime_checkable
@@ -296,6 +437,10 @@ __all__ = [
     "CAPABILITY_CONNECTORS_TABULAR_SOURCES",
     "DEFAULT_PREVIEW_BYTES",
     "DEFAULT_PREVIEW_TIMEOUT_MS",
+    "CsvValueMode",
+    "TabularCsvSource",
+    "tabular_type_name",
+    "infer_tabular_schema",
     "TabularColumn",
     "TabularPreviewDiagnostic",
     "TabularPushdown",
@@ -313,6 +458,10 @@ __all__ = [
     "TabularSourceUnavailableError",
     "TabularSourceValidationError",
     "parse_tabular_csv",
+    "csv_source_payload",
+    "csv_source_summary",
+    "verified_csv_source_text",
+    "csv_projection_matches",
     "tabular_snapshot_writer",
     "tabular_source_provider",
 ]

src/govoplan_core/core/tasks.py

@@ -0,0 +1,332 @@
+from __future__ import annotations
+
+from collections.abc import Callable, Mapping
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Literal, Protocol, runtime_checkable
+
+from govoplan_core.core.modules import ModuleContext
+
+
+WorkItemStatus = Literal[
+    "open",
+    "in_progress",
+    "deferred",
+    "blocked",
+    "completed",
+    "cancelled",
+]
+WorkItemPriority = Literal["low", "normal", "high", "urgent"]
+WorkAssignmentKind = Literal[
+    "account",
+    "group",
+    "role",
+    "function",
+    "function_assignment",
+    "anyone",
+]
+WORK_ITEM_CONTRACT_VERSION = "1"
+CAPABILITY_TASK_COMMANDS = "tasks.commands"
+
+
+@dataclass(frozen=True, slots=True)
+class WorkAssignmentRef:
+    kind: WorkAssignmentKind
+    id: str
+    label: str | None = None
+
+    def __post_init__(self) -> None:
+        if self.kind not in {
+            "account",
+            "group",
+            "role",
+            "function",
+            "function_assignment",
+            "anyone",
+        }:
+            raise ValueError(f"Unsupported work-assignment kind: {self.kind!r}.")
+        if not self.id.strip():
+            raise ValueError("Work assignments require an id.")
+        if len(self.id) > 255:
+            raise ValueError("Work assignment ids are limited to 255 characters.")
+        if self.label is not None and len(self.label) > 500:
+            raise ValueError("Work assignment labels are limited to 500 characters.")
+        if self.kind == "anyone" and self.id != "*":
+            raise ValueError("Broad work assignments use the canonical '*' id.")
+
+
+@dataclass(frozen=True, slots=True)
+class WorkSourceRef:
+    module_id: str
+    resource_type: str
+    resource_id: str
+    revision: str | None = None
+    url: str | None = None
+    label: str | None = None
+
+    def __post_init__(self) -> None:
+        required = (self.module_id, self.resource_type, self.resource_id)
+        if any(not value.strip() for value in required):
+            raise ValueError("Work source references require module, type, and id.")
+        limits = {
+            "module_id": 100,
+            "resource_type": 100,
+            "resource_id": 255,
+            "revision": 255,
+            "url": 1_500,
+            "label": 500,
+        }
+        for field_name, limit in limits.items():
+            value = getattr(self, field_name)
+            if value is not None and len(value) > limit:
+                raise ValueError(
+                    f"Work source {field_name} is limited to {limit} characters."
+                )
+
+
+@dataclass(frozen=True, slots=True)
+class WorkItem:
+    id: str
+    provider_id: str
+    owner_module: str
+    tenant_id: str
+    title: str
+    status: WorkItemStatus = "open"
+    priority: WorkItemPriority = "normal"
+    summary: str | None = None
+    required_action: str | None = None
+    action_url: str | None = None
+    due_at: datetime | None = None
+    deferred_until: datetime | None = None
+    assignments: tuple[WorkAssignmentRef, ...] = ()
+    sources: tuple[WorkSourceRef, ...] = ()
+    provenance: Mapping[str, object] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
+    revision: str = "1"
+    created_at: datetime | None = None
+    updated_at: datetime | None = None
+
+    def __post_init__(self) -> None:
+        required = {
+            "id": self.id,
+            "provider_id": self.provider_id,
+            "owner_module": self.owner_module,
+            "tenant_id": self.tenant_id,
+            "title": self.title,
+            "revision": self.revision,
+        }
+        if any(not value.strip() for value in required.values()):
+            raise ValueError(
+                "Work items require stable identity, owner, tenant, and title."
+            )
+        limits = {
+            "id": 255,
+            "provider_id": 200,
+            "owner_module": 100,
+            "tenant_id": 255,
+            "title": 500,
+            "summary": 4_000,
+            "required_action": 500,
+            "action_url": 1_500,
+            "revision": 255,
+        }
+        for field_name, limit in limits.items():
+            value = getattr(self, field_name)
+            if value is not None and len(value) > limit:
+                raise ValueError(
+                    f"Work item {field_name} is limited to {limit} characters."
+                )
+        if len(self.assignments) > 100 or len(self.sources) > 100:
+            raise ValueError("Work items support at most 100 assignments and sources.")
+        _validate_action_url(self.action_url)
+        if self.status not in {
+            "open",
+            "in_progress",
+            "deferred",
+            "blocked",
+            "completed",
+            "cancelled",
+        }:
+            raise ValueError(f"Unsupported work-item status: {self.status!r}.")
+        if self.priority not in {"low", "normal", "high", "urgent"}:
+            raise ValueError(f"Unsupported work-item priority: {self.priority!r}.")
+
+
+@dataclass(frozen=True, slots=True)
+class WorkItemQuery:
+    tenant_id: str
+    statuses: tuple[WorkItemStatus, ...] = (
+        "open",
+        "in_progress",
+        "deferred",
+        "blocked",
+    )
+    priorities: tuple[WorkItemPriority, ...] = ()
+    provider_ids: tuple[str, ...] = ()
+    owner_modules: tuple[str, ...] = ()
+    due_before: datetime | None = None
+    text: str = ""
+    limit: int = 100
+
+    def __post_init__(self) -> None:
+        if not self.tenant_id.strip():
+            raise ValueError("Work-item queries require a tenant.")
+        if not 1 <= self.limit <= 500:
+            raise ValueError("Work-item query limits must be between 1 and 500.")
+        normalized = self.text.strip()
+        if len(normalized) > 500:
+            raise ValueError("Work-item query text is limited to 500 characters.")
+        if len(self.provider_ids) > 50 or len(self.owner_modules) > 50:
+            raise ValueError("Work-item queries support at most 50 provider filters.")
+        object.__setattr__(self, "text", normalized)
+
+
+@dataclass(frozen=True, slots=True)
+class WorkItemPage:
+    items: tuple[WorkItem, ...]
+    total: int
+    truncated: bool = False
+
+    def __post_init__(self) -> None:
+        if self.total < len(self.items):
+            raise ValueError(
+                "Work-item totals cannot be smaller than the returned page."
+            )
+
+
+@runtime_checkable
+class WorkItemProvider(Protocol):
+    def list_items(
+        self,
+        session: object,
+        principal: object,
+        *,
+        query: WorkItemQuery,
+    ) -> WorkItemPage:
+        """Return only items the current principal may discover and act on."""
+
+
+@dataclass(frozen=True, slots=True)
+class TaskCreateCommand:
+    tenant_id: str
+    title: str
+    idempotency_key: str
+    summary: str | None = None
+    priority: WorkItemPriority = "normal"
+    due_at: datetime | None = None
+    required_action: str | None = None
+    action_url: str | None = None
+    assignments: tuple[WorkAssignmentRef, ...] = ()
+    sources: tuple[WorkSourceRef, ...] = ()
+    provenance: Mapping[str, object] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        if not self.tenant_id.strip() or not self.title.strip():
+            raise ValueError("Task commands require a tenant and title.")
+        if not self.idempotency_key.strip() or len(self.idempotency_key) > 255:
+            raise ValueError("Task commands require a bounded idempotency key.")
+        if not self.assignments:
+            raise ValueError("Explicit tasks require at least one assignment.")
+        if self.priority not in {"low", "normal", "high", "urgent"}:
+            raise ValueError(f"Unsupported task priority: {self.priority!r}.")
+        limits = {
+            "title": 500,
+            "summary": 4_000,
+            "required_action": 500,
+            "action_url": 1_500,
+        }
+        for field_name, limit in limits.items():
+            value = getattr(self, field_name)
+            if value is not None and len(value) > limit:
+                raise ValueError(
+                    f"Task command {field_name} is limited to {limit} characters."
+                )
+        if len(self.assignments) > 100 or len(self.sources) > 100:
+            raise ValueError(
+                "Task commands support at most 100 assignments and sources."
+            )
+        _validate_action_url(self.action_url)
+
+
+def _validate_action_url(value: str | None) -> None:
+    if value is None:
+        return
+    candidate = value.strip()
+    if not candidate:
+        return
+    if (
+        not candidate.startswith("/")
+        or candidate.startswith("//")
+        or "\\" in candidate
+        or any(ord(character) < 32 or ord(character) == 127 for character in candidate)
+    ):
+        raise ValueError("Work-item action URLs must be application-relative paths.")
+
+
+@runtime_checkable
+class TaskCommandProvider(Protocol):
+    def create_task(
+        self,
+        session: object,
+        principal: object,
+        *,
+        command: TaskCreateCommand,
+    ) -> WorkItem: ...
+
+
+WorkItemProviderFactory = Callable[[ModuleContext], WorkItemProvider]
+
+
+@dataclass(frozen=True, slots=True)
+class WorkItemProviderRegistration:
+    id: str
+    factory: WorkItemProviderFactory
+    order: int = 100
+
+    def create(self, context: ModuleContext) -> WorkItemProvider:
+        provider = self.factory(context)
+        if not isinstance(provider, WorkItemProvider):
+            raise TypeError(
+                f"Work-item provider {self.id!r} does not implement WorkItemProvider."
+            )
+        return provider
+
+
+@dataclass(frozen=True, slots=True)
+class RegisteredWorkItemProvider:
+    module_id: str
+    registration: WorkItemProviderRegistration
+
+
+def task_command_provider(registry: object | None) -> TaskCommandProvider | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not hasattr(registry, "capability")
+        or not registry.has_capability(CAPABILITY_TASK_COMMANDS)
+    ):
+        return None
+    provider = registry.capability(CAPABILITY_TASK_COMMANDS)
+    return provider if isinstance(provider, TaskCommandProvider) else None
+
+
+__all__ = [
+    "CAPABILITY_TASK_COMMANDS",
+    "RegisteredWorkItemProvider",
+    "TaskCommandProvider",
+    "TaskCreateCommand",
+    "WORK_ITEM_CONTRACT_VERSION",
+    "WorkAssignmentKind",
+    "WorkAssignmentRef",
+    "WorkItem",
+    "WorkItemPage",
+    "WorkItemPriority",
+    "WorkItemProvider",
+    "WorkItemProviderFactory",
+    "WorkItemProviderRegistration",
+    "WorkItemQuery",
+    "WorkItemStatus",
+    "WorkSourceRef",
+    "task_command_provider",
+]

src/govoplan_core/core/templates.py

@@ -8,6 +8,7 @@ from typing import Literal, Protocol, runtime_checkable
 
 CAPABILITY_TEMPLATE_CATALOG = "templates.catalog"
 CAPABILITY_TEMPLATE_RENDERER = "templates.renderer"
+CAPABILITY_TEMPLATE_CONTENT_LIBRARY = "templates.content_library"
 
 TemplateType = Literal[
     "label",
@@ -17,6 +18,7 @@ TemplateType = Literal[
     "form_letter",
     "list_layout",
     "email",
+    "content_fragment",
     "generic",
 ]
 TemplateOutputFormat = Literal["html", "text"]
@@ -79,6 +81,10 @@ class TemplateRevisionRef:
     locale: str
     required_fields: tuple[TemplateFieldRequirement, ...]
     output_profiles: tuple[TemplateOutputProfile, ...]
+    content_text: str | None = None
+    content_html: str | None = None
+    layout: Mapping[str, object] = field(default_factory=dict)
+    metadata: Mapping[str, object] = field(default_factory=dict)
     published_at: datetime | None = None
     provenance: Mapping[str, object] = field(default_factory=dict)
 
@@ -166,6 +172,23 @@ class TemplateRenderResult:
     payload: bytes | None = None
 
 
+@dataclass(frozen=True, slots=True)
+class TemplateContentDraftRequest:
+    """Provider-neutral request for a reusable text/HTML content draft."""
+
+    name: str
+    template_type: TemplateType
+    usages: tuple[str, ...]
+    content_text: str | None = None
+    content_html: str | None = None
+    description: str | None = None
+    locale: str = "de"
+    scope_type: Literal["tenant", "group", "user"] = "tenant"
+    scope_id: str | None = None
+    metadata: Mapping[str, object] = field(default_factory=dict)
+    required_fields: tuple[TemplateFieldRequirement, ...] = ()
+
+
 @runtime_checkable
 class TemplateCatalogProvider(Protocol):
     def list_templates(
@@ -213,13 +236,29 @@ class TemplateRendererProvider(Protocol):
     ) -> TemplateRenderResult: ...
 
 
+@runtime_checkable
+class TemplateContentLibraryProvider(Protocol):
+    """Create reusable content drafts while Templates retains ownership."""
+
+    def create_content_draft(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: TemplateContentDraftRequest,
+    ) -> TemplateRef: ...
+
+
 __all__ = [
     "CAPABILITY_TEMPLATE_CATALOG",
+    "CAPABILITY_TEMPLATE_CONTENT_LIBRARY",
     "CAPABILITY_TEMPLATE_RENDERER",
     "TemplateArtifactRef",
     "TemplateCatalogProvider",
     "TemplateCompatibility",
     "TemplateCompatibilityError",
+    "TemplateContentDraftRequest",
+    "TemplateContentLibraryProvider",
     "TemplateContractError",
     "TemplateFieldRequirement",
     "TemplateNotFoundError",

src/govoplan_core/core/temporal.py

@@ -0,0 +1,182 @@
+from __future__ import annotations
+
+from contextvars import ContextVar, Token
+from dataclasses import dataclass, field
+from datetime import UTC, datetime
+from typing import Literal
+
+
+TemporalValidityMode = Literal["current", "at", "all"]
+VALIDITY_MODES = frozenset({"current", "at", "all"})
+
+VALIDITY_MODE_HEADER = "X-Govoplan-Validity-Mode"
+VALID_AT_HEADER = "X-Govoplan-Valid-At"
+RECORDED_AT_HEADER = "X-Govoplan-Recorded-At"
+TEMPORAL_EVALUATED_AT_HEADER = "X-Govoplan-Temporal-Evaluated-At"
+TEMPORAL_VARY_HEADERS = (
+    VALIDITY_MODE_HEADER,
+    VALID_AT_HEADER,
+    RECORDED_AT_HEADER,
+)
+
+
+class TemporalContextError(ValueError):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class TemporalDataContext:
+    """Bitemporal read context.
+
+    Valid time answers when a fact applies. Recorded time answers which version
+    of that fact was known to the system. Authorization remains outside this
+    context and is always evaluated under the current security state.
+    """
+
+    validity_mode: TemporalValidityMode = "current"
+    valid_at: datetime | None = None
+    recorded_at: datetime | None = None
+    evaluated_at: datetime = field(default_factory=lambda: datetime.now(UTC))
+
+    def __post_init__(self) -> None:
+        if self.validity_mode not in VALIDITY_MODES:
+            raise TemporalContextError(
+                f"Unsupported temporal validity mode: {self.validity_mode!r}."
+            )
+        for name in ("valid_at", "recorded_at", "evaluated_at"):
+            value = getattr(self, name)
+            if value is not None and value.tzinfo is None:
+                raise TemporalContextError(f"Temporal {name} must include a timezone.")
+        if self.validity_mode == "at" and self.valid_at is None:
+            raise TemporalContextError("Validity mode 'at' requires valid_at.")
+        if self.validity_mode != "at" and self.valid_at is not None:
+            raise TemporalContextError(
+                "valid_at is only permitted when validity mode is 'at'."
+            )
+
+    @property
+    def validity_instant(self) -> datetime | None:
+        if self.validity_mode == "all":
+            return None
+        if self.validity_mode == "at":
+            return self.valid_at
+        return self.evaluated_at
+
+    @property
+    def is_default(self) -> bool:
+        return self.validity_mode == "current" and self.recorded_at is None
+
+    def to_dict(self) -> dict[str, str | None]:
+        return {
+            "validity_mode": self.validity_mode,
+            "valid_at": _datetime_text(self.valid_at),
+            "recorded_at": _datetime_text(self.recorded_at),
+            "evaluated_at": _datetime_text(self.evaluated_at),
+        }
+
+
+_temporal_context: ContextVar[TemporalDataContext | None] = ContextVar(
+    "govoplan_temporal_data_context",
+    default=None,
+)
+
+
+def parse_temporal_data_context(
+    *,
+    validity_mode: str | None = None,
+    valid_at: str | None = None,
+    recorded_at: str | None = None,
+    evaluated_at: datetime | None = None,
+) -> TemporalDataContext:
+    clean_mode = (validity_mode or "current").strip().lower()
+    if clean_mode not in VALIDITY_MODES:
+        raise TemporalContextError(
+            "Temporal validity mode must be one of: current, at, all."
+        )
+    return TemporalDataContext(
+        validity_mode=clean_mode,  # type: ignore[arg-type]
+        valid_at=_parse_datetime(valid_at, "valid_at"),
+        recorded_at=_parse_datetime(recorded_at, "recorded_at"),
+        evaluated_at=evaluated_at or datetime.now(UTC),
+    )
+
+
+def current_temporal_data_context() -> TemporalDataContext:
+    return _temporal_context.get() or TemporalDataContext()
+
+
+def bind_temporal_data_context(
+    context: TemporalDataContext,
+) -> Token[TemporalDataContext | None]:
+    return _temporal_context.set(context)
+
+
+def reset_temporal_data_context(token: Token[TemporalDataContext | None]) -> None:
+    _temporal_context.reset(token)
+
+
+def temporal_revision_matches(
+    context: TemporalDataContext,
+    *,
+    valid_from: datetime | None = None,
+    valid_to: datetime | None = None,
+    revision_recorded_at: datetime | None = None,
+    superseded_at: datetime | None = None,
+) -> bool:
+    cutoff = context.recorded_at
+    if cutoff is None:
+        if superseded_at is not None:
+            return False
+    else:
+        if revision_recorded_at is None or revision_recorded_at > cutoff:
+            return False
+        if superseded_at is not None and superseded_at <= cutoff:
+            return False
+
+    instant = context.validity_instant
+    if instant is None:
+        return True
+    return (valid_from is None or valid_from <= instant) and (
+        valid_to is None or valid_to > instant
+    )
+
+
+def _parse_datetime(value: str | None, name: str) -> datetime | None:
+    clean = str(value or "").strip()
+    if not clean:
+        return None
+    if len(clean) > 64:
+        raise TemporalContextError(f"Temporal {name} is too long.")
+    normalized = f"{clean[:-1]}+00:00" if clean.endswith(("Z", "z")) else clean
+    try:
+        parsed = datetime.fromisoformat(normalized)
+    except ValueError as exc:
+        raise TemporalContextError(
+            f"Temporal {name} must be an ISO 8601 timestamp."
+        ) from exc
+    if parsed.tzinfo is None:
+        raise TemporalContextError(f"Temporal {name} must include a timezone.")
+    return parsed.astimezone(UTC)
+
+
+def _datetime_text(value: datetime | None) -> str | None:
+    if value is None:
+        return None
+    return value.astimezone(UTC).isoformat().replace("+00:00", "Z")
+
+
+__all__ = [
+    "RECORDED_AT_HEADER",
+    "TEMPORAL_EVALUATED_AT_HEADER",
+    "TEMPORAL_VARY_HEADERS",
+    "VALIDITY_MODE_HEADER",
+    "VALID_AT_HEADER",
+    "TemporalContextError",
+    "TemporalDataContext",
+    "TemporalValidityMode",
+    "bind_temporal_data_context",
+    "current_temporal_data_context",
+    "parse_temporal_data_context",
+    "reset_temporal_data_context",
+    "temporal_revision_matches",
+]

src/govoplan_core/core/tenant_erasure.py

@@ -0,0 +1,467 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass, field
+from datetime import UTC, datetime
+from typing import Literal, Protocol, runtime_checkable
+
+
+TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX = "tenancy.erasure_provider."
+
+TenantErasureDisposition = Literal[
+    "erase",
+    "retain",
+    "legal_hold",
+    "external_cleanup",
+    "key_destroy",
+    "backup_expiry",
+    "unavailable",
+]
+TenantErasureStepKind = Literal[
+    "export",
+    "erase",
+    "retain",
+    "external_cleanup",
+    "key_destroy",
+    "backup_expiry",
+    "verify",
+]
+TenantErasureResultState = Literal[
+    "completed",
+    "pending",
+    "blocked",
+    "outcome_unknown",
+]
+
+_DISPOSITIONS = frozenset(
+    {
+        "erase",
+        "retain",
+        "legal_hold",
+        "external_cleanup",
+        "key_destroy",
+        "backup_expiry",
+        "unavailable",
+    }
+)
+_STEP_KINDS = frozenset(
+    {
+        "export",
+        "erase",
+        "retain",
+        "external_cleanup",
+        "key_destroy",
+        "backup_expiry",
+        "verify",
+    }
+)
+_RESULT_STATES = frozenset(
+    {"completed", "pending", "blocked", "outcome_unknown"}
+)
+
+
+def _text(value: str, label: str, *, maximum: int) -> str:
+    normalized = value.strip()
+    if (
+        not normalized
+        or len(normalized) > maximum
+        or any(ord(character) < 32 for character in normalized)
+    ):
+        raise ValueError(f"Tenant erasure {label} is invalid.")
+    return normalized
+
+
+def _texts(
+    values: tuple[str, ...],
+    label: str,
+    *,
+    maximum_items: int = 100,
+    maximum_length: int = 500,
+) -> tuple[str, ...]:
+    if len(values) > maximum_items:
+        raise ValueError(f"Tenant erasure {label} has too many entries.")
+    normalized = tuple(
+        _text(value, label, maximum=maximum_length) for value in values
+    )
+    if len(normalized) != len(set(normalized)):
+        raise ValueError(f"Tenant erasure {label} contains duplicates.")
+    return normalized
+
+
+def _metrics(values: Mapping[str, int]) -> dict[str, int]:
+    if len(values) > 30:
+        raise ValueError("Tenant erasure metrics has too many entries.")
+    normalized: dict[str, int] = {}
+    for key, value in values.items():
+        normalized_key = _text(key, "metric key", maximum=80)
+        if type(value) is not int or value < 0:
+            raise ValueError("Tenant erasure metric values must be non-negative integers.")
+        normalized[normalized_key] = value
+    return normalized
+
+
+@dataclass(frozen=True, slots=True)
+class TenantErasureResource:
+    resource_type: str
+    count: int
+    disposition: TenantErasureDisposition
+    summary: str
+    governance_ref: str | None = None
+    external: bool = False
+
+    def __post_init__(self) -> None:
+        _text(self.resource_type, "resource type", maximum=120)
+        _text(self.summary, "resource summary", maximum=1000)
+        if type(self.count) is not int or self.count < 0:
+            raise ValueError("Tenant erasure resource count is invalid.")
+        if self.disposition not in _DISPOSITIONS:
+            raise ValueError("Tenant erasure resource disposition is invalid.")
+        if self.governance_ref is not None:
+            _text(self.governance_ref, "governance reference", maximum=300)
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "resource_type": self.resource_type,
+            "count": self.count,
+            "disposition": self.disposition,
+            "summary": self.summary,
+            "governance_ref": self.governance_ref,
+            "external": self.external,
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class TenantErasureStep:
+    step_id: str
+    kind: TenantErasureStepKind
+    summary: str
+    destructive: bool
+    irreversible: bool
+    requires_reconciliation: bool = False
+    depends_on: tuple[str, ...] = ()
+
+    def __post_init__(self) -> None:
+        _text(self.step_id, "step id", maximum=160)
+        _text(self.summary, "step summary", maximum=1000)
+        if self.kind not in _STEP_KINDS:
+            raise ValueError("Tenant erasure step kind is invalid.")
+        _texts(self.depends_on, "step dependencies", maximum_length=160)
+        if self.step_id in self.depends_on:
+            raise ValueError("Tenant erasure step cannot depend on itself.")
+        if self.irreversible and not self.destructive:
+            raise ValueError("An irreversible tenant erasure step must be destructive.")
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "step_id": self.step_id,
+            "kind": self.kind,
+            "summary": self.summary,
+            "destructive": self.destructive,
+            "irreversible": self.irreversible,
+            "requires_reconciliation": self.requires_reconciliation,
+            "depends_on": list(self.depends_on),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class TenantErasurePreview:
+    module_id: str
+    complete: bool
+    resources: tuple[TenantErasureResource, ...] = ()
+    steps: tuple[TenantErasureStep, ...] = ()
+    blockers: tuple[str, ...] = ()
+    warnings: tuple[str, ...] = ()
+    provider_revision: str = "1"
+
+    def __post_init__(self) -> None:
+        _text(self.module_id, "module id", maximum=120)
+        _text(self.provider_revision, "provider revision", maximum=120)
+        _texts(self.blockers, "blockers", maximum_length=1000)
+        _texts(self.warnings, "warnings", maximum_length=1000)
+        if len(self.resources) > 500 or len(self.steps) > 500:
+            raise ValueError("Tenant erasure preview is too large.")
+        resource_types = [item.resource_type for item in self.resources]
+        if len(resource_types) != len(set(resource_types)):
+            raise ValueError("Tenant erasure preview repeats a resource type.")
+        resources_requiring_action = tuple(
+            item for item in self.resources if item.count > 0
+        )
+        if resources_requiring_action and not self.steps and not self.blockers:
+            raise ValueError(
+                "Tenant erasure resources require steps or an explicit blocker."
+            )
+        if any(
+            item.count > 0 and item.disposition == "unavailable"
+            for item in self.resources
+        ) and not self.blockers:
+            raise ValueError(
+                "Unavailable tenant erasure resources require an explicit blocker."
+            )
+        if not self.complete and not self.blockers:
+            raise ValueError(
+                "An incomplete tenant erasure preview requires an explicit blocker."
+            )
+        step_ids = [item.step_id for item in self.steps]
+        if len(step_ids) != len(set(step_ids)):
+            raise ValueError("Tenant erasure preview repeats a step id.")
+        known_step_ids = set(step_ids)
+        if any(
+            dependency not in known_step_ids
+            for step in self.steps
+            for dependency in step.depends_on
+        ):
+            raise ValueError("Tenant erasure step references an unknown dependency.")
+        remaining = {
+            step.step_id: set(step.depends_on)
+            for step in self.steps
+        }
+        resolved: set[str] = set()
+        while remaining:
+            ready = sorted(
+                step_id
+                for step_id, dependencies in remaining.items()
+                if dependencies.issubset(resolved)
+            )
+            if not ready:
+                raise ValueError("Tenant erasure step dependencies contain a cycle.")
+            resolved.update(ready)
+            for step_id in ready:
+                remaining.pop(step_id)
+
+    @property
+    def allowed(self) -> bool:
+        return self.complete and not self.blockers
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "module_id": self.module_id,
+            "complete": self.complete,
+            "allowed": self.allowed,
+            "provider_revision": self.provider_revision,
+            "resources": [item.to_dict() for item in self.resources],
+            "steps": [item.to_dict() for item in self.steps],
+            "blockers": list(self.blockers),
+            "warnings": list(self.warnings),
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class TenantErasureStepResult:
+    state: TenantErasureResultState
+    summary: str
+    receipt_ref: str | None = None
+    metrics: Mapping[str, int] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        if self.state not in _RESULT_STATES:
+            raise ValueError("Tenant erasure result state is invalid.")
+        _text(self.summary, "result summary", maximum=1000)
+        if self.receipt_ref is not None:
+            _text(self.receipt_ref, "receipt reference", maximum=500)
+        _metrics(self.metrics)
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "state": self.state,
+            "summary": self.summary,
+            "receipt_ref": self.receipt_ref,
+            "metrics": dict(sorted(_metrics(self.metrics).items())),
+        }
+
+
+@runtime_checkable
+class TenantErasureProvider(Protocol):
+    module_id: str
+
+    def preview_tenant_erasure(
+        self,
+        session: object,
+        tenant_id: str,
+    ) -> TenantErasurePreview:
+        ...
+
+    def execute_tenant_erasure_step(
+        self,
+        session: object,
+        tenant_id: str,
+        step_id: str,
+        idempotency_key: str,
+    ) -> TenantErasureStepResult:
+        ...
+
+    def reconcile_tenant_erasure_step(
+        self,
+        session: object,
+        tenant_id: str,
+        step_id: str,
+        idempotency_key: str,
+    ) -> TenantErasureStepResult:
+        ...
+
+
+@dataclass(frozen=True, slots=True)
+class TenantErasureInventory:
+    tenant_id: str
+    generated_at: datetime
+    complete: bool
+    modules: tuple[TenantErasurePreview, ...]
+
+    @property
+    def allowed(self) -> bool:
+        return self.complete and all(item.allowed for item in self.modules)
+
+    def to_dict(self) -> dict[str, object]:
+        generated_at = self.generated_at
+        if generated_at.tzinfo is None:
+            generated_at = generated_at.replace(tzinfo=UTC)
+        return {
+            "schema_version": 1,
+            "tenant_id": self.tenant_id,
+            "generated_at": generated_at.astimezone(UTC).isoformat(),
+            "complete": self.complete,
+            "allowed": self.allowed,
+            "modules": [item.to_dict() for item in self.modules],
+        }
+
+
+def tenant_erasure_providers(registry: object) -> dict[str, TenantErasureProvider]:
+    capability_names = getattr(registry, "capability_names", None)
+    capability = getattr(registry, "capability", None)
+    if not callable(capability_names) or not callable(capability):
+        raise ValueError("Tenant erasure requires a module registry.")
+    providers: dict[str, TenantErasureProvider] = {}
+    for capability_name in sorted(capability_names()):
+        if not capability_name.startswith(TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX):
+            continue
+        expected_module_id = capability_name.removeprefix(
+            TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX
+        )
+        provider = capability(capability_name)
+        if not isinstance(provider, TenantErasureProvider):
+            raise TypeError(
+                f"Tenant erasure provider {expected_module_id or 'unknown'} is invalid."
+            )
+        module_id = _text(provider.module_id, "provider module id", maximum=120)
+        if module_id != expected_module_id or module_id in providers:
+            raise ValueError("Tenant erasure provider identity is invalid.")
+        providers[module_id] = provider
+    return providers
+
+
+def collect_tenant_erasure_inventory(
+    registry: object,
+    session: object,
+    tenant_id: str,
+    *,
+    observed_at: datetime | None = None,
+) -> TenantErasureInventory:
+    normalized_tenant_id = _text(tenant_id, "tenant id", maximum=120)
+    manifests = getattr(registry, "manifests", None)
+    summary_providers = getattr(registry, "tenant_summary_providers", None)
+    if not callable(manifests) or not callable(summary_providers):
+        raise ValueError("Tenant erasure inventory requires a module registry.")
+    provider_by_module = tenant_erasure_providers(registry)
+    summary_by_module = dict(summary_providers())
+    manifest_ids = {
+        str(manifest.id)
+        for manifest in manifests()
+        if getattr(manifest, "id", None)
+    }
+    module_ids = manifest_ids | set(summary_by_module) | set(provider_by_module)
+    previews: list[TenantErasurePreview] = []
+    complete = True
+    for module_id in sorted(module_ids):
+        provider = provider_by_module.get(module_id)
+        if provider is not None:
+            try:
+                preview = provider.preview_tenant_erasure(session, normalized_tenant_id)
+                if not isinstance(preview, TenantErasurePreview):
+                    raise TypeError("provider returned an invalid preview")
+                if preview.module_id != module_id:
+                    raise ValueError("provider returned another module's preview")
+            except Exception as exc:
+                complete = False
+                preview = TenantErasurePreview(
+                    module_id=module_id,
+                    complete=False,
+                    blockers=(
+                        f"{type(exc).__name__}: provider preview could not be completed",
+                    ),
+                )
+            previews.append(preview)
+            complete = complete and preview.complete
+            continue
+        summary_provider = summary_by_module.get(module_id)
+        if summary_provider is None:
+            previews.append(
+                TenantErasurePreview(
+                    module_id=module_id,
+                    complete=True,
+                    warnings=(
+                        "Module declares no tenant-owned summary or erasure provider; no tenant persistence is in scope.",
+                    ),
+                    provider_revision="manifest-no-tenant-data",
+                )
+            )
+            continue
+        try:
+            raw_counts = summary_provider(session, normalized_tenant_id)
+            counts = _metrics({str(key): int(value) for key, value in raw_counts.items()})
+            resources = tuple(
+                TenantErasureResource(
+                    resource_type=resource_type,
+                    count=count,
+                    disposition="unavailable" if count else "erase",
+                    summary=(
+                        "Tenant-owned data requires a module erasure provider."
+                        if count
+                        else "The module reported no tenant-owned records."
+                    ),
+                )
+                for resource_type, count in sorted(counts.items())
+            )
+            blockers = (
+                ("Tenant-owned data exists but the module has no erasure provider.",)
+                if any(counts.values())
+                else ()
+            )
+            preview = TenantErasurePreview(
+                module_id=module_id,
+                complete=True,
+                resources=resources,
+                blockers=blockers,
+                provider_revision="tenant-summary-fallback",
+            )
+        except Exception as exc:
+            complete = False
+            preview = TenantErasurePreview(
+                module_id=module_id,
+                complete=False,
+                blockers=(
+                    f"{type(exc).__name__}: tenant summary could not be completed",
+                ),
+                provider_revision="tenant-summary-fallback",
+            )
+        previews.append(preview)
+    timestamp = observed_at or datetime.now(UTC)
+    if timestamp.tzinfo is None:
+        timestamp = timestamp.replace(tzinfo=UTC)
+    return TenantErasureInventory(
+        tenant_id=normalized_tenant_id,
+        generated_at=timestamp,
+        complete=complete,
+        modules=tuple(previews),
+    )
+
+
+__all__ = [
+    "TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX",
+    "TenantErasureInventory",
+    "TenantErasurePreview",
+    "TenantErasureProvider",
+    "TenantErasureResource",
+    "TenantErasureStep",
+    "TenantErasureStepResult",
+    "collect_tenant_erasure_inventory",
+    "tenant_erasure_providers",
+]

src/govoplan_core/core/tickets.py

@@ -0,0 +1,179 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Mapping, Protocol, runtime_checkable
+
+
+TICKET_INTEGRATION_CONTRACT_VERSION = "1"
+CAPABILITY_TICKET_ROUTING = "tickets.routing"
+CAPABILITY_TICKET_CASE_ESCALATION = "tickets.case_escalation"
+
+
+@dataclass(frozen=True, slots=True)
+class TicketRoutingRequest:
+    tenant_id: str
+    ticket_id: str
+    ticket_type: str
+    priority: str
+    title: str
+    received_at: datetime
+    queue_hint: str | None = None
+    attributes: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _required(self.tenant_id, "Ticket routing tenant", 255)
+        _required(self.ticket_id, "Ticket routing ticket", 255)
+        _required(self.ticket_type, "Ticket routing type", 80)
+        _required(self.priority, "Ticket routing priority", 40)
+        _required(self.title, "Ticket routing title", 500)
+        _aware(self.received_at, "Ticket routing received_at")
+        _optional(self.queue_hint, "Ticket routing queue hint", 255)
+        if len(self.attributes) > 100:
+            raise ValueError("Ticket routing attributes are limited to 100 entries.")
+
+
+@dataclass(frozen=True, slots=True)
+class TicketRoutingPlan:
+    provider_id: str
+    queue_ref: str | None = None
+    service_target_at: datetime | None = None
+    explanation: str | None = None
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _required(self.provider_id, "Ticket routing provider", 200)
+        _optional(self.queue_ref, "Ticket routing queue reference", 255)
+        _optional(self.explanation, "Ticket routing explanation", 4_000)
+        _aware(self.service_target_at, "Ticket routing service_target_at")
+        if len(self.metadata) > 100:
+            raise ValueError("Ticket routing metadata is limited to 100 entries.")
+
+
+@runtime_checkable
+class TicketRoutingProvider(Protocol):
+    def route_ticket(
+        self,
+        session: object,
+        principal: object,
+        *,
+        request: TicketRoutingRequest,
+    ) -> TicketRoutingPlan: ...
+
+
+@dataclass(frozen=True, slots=True)
+class TicketCaseEscalationCommand:
+    tenant_id: str
+    ticket_id: str
+    ticket_number: str
+    title: str
+    case_type_key: str
+    occurred_at: datetime
+    idempotency_key: str
+    handoff_note: str | None = None
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _required(self.tenant_id, "Ticket escalation tenant", 255)
+        _required(self.ticket_id, "Ticket escalation ticket", 255)
+        _required(self.ticket_number, "Ticket escalation number", 255)
+        _required(self.title, "Ticket escalation title", 500)
+        _required(self.case_type_key, "Ticket escalation case type", 120)
+        _required(self.idempotency_key, "Ticket escalation idempotency key", 255)
+        _optional(self.handoff_note, "Ticket escalation handoff note", 10_000)
+        _aware(self.occurred_at, "Ticket escalation occurred_at")
+        if len(self.metadata) > 100:
+            raise ValueError("Ticket escalation metadata is limited to 100 entries.")
+
+
+@dataclass(frozen=True, slots=True)
+class TicketCaseEscalationResult:
+    provider_id: str
+    case_id: str
+    case_number: str
+    case_url: str
+    replayed: bool = False
+    metadata: Mapping[str, object] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        _required(self.provider_id, "Ticket escalation provider", 200)
+        _required(self.case_id, "Ticket escalation case", 255)
+        _required(self.case_number, "Ticket escalation case number", 255)
+        _relative_url(self.case_url)
+        if len(self.metadata) > 100:
+            raise ValueError("Ticket escalation metadata is limited to 100 entries.")
+
+
+@runtime_checkable
+class TicketCaseEscalationProvider(Protocol):
+    def escalate_ticket(
+        self,
+        session: object,
+        principal: object,
+        *,
+        command: TicketCaseEscalationCommand,
+    ) -> TicketCaseEscalationResult: ...
+
+
+def ticket_routing_provider(registry: object | None) -> TicketRoutingProvider | None:
+    provider = _capability(registry, CAPABILITY_TICKET_ROUTING)
+    return provider if isinstance(provider, TicketRoutingProvider) else None
+
+
+def ticket_case_escalation_provider(
+    registry: object | None,
+) -> TicketCaseEscalationProvider | None:
+    provider = _capability(registry, CAPABILITY_TICKET_CASE_ESCALATION)
+    return provider if isinstance(provider, TicketCaseEscalationProvider) else None
+
+
+def _capability(registry: object | None, name: str) -> object | None:
+    if (
+        registry is None
+        or not hasattr(registry, "has_capability")
+        or not hasattr(registry, "capability")
+        or not registry.has_capability(name)
+    ):
+        return None
+    return registry.capability(name)
+
+
+def _required(value: str, label: str, maximum: int) -> None:
+    if not value.strip() or len(value) > maximum:
+        raise ValueError(f"{label} must contain 1 to {maximum} characters.")
+
+
+def _optional(value: str | None, label: str, maximum: int) -> None:
+    if value is not None and (not value.strip() or len(value) > maximum):
+        raise ValueError(f"{label} must contain 1 to {maximum} characters when set.")
+
+
+def _aware(value: datetime | None, label: str) -> None:
+    if value is not None and (value.tzinfo is None or value.utcoffset() is None):
+        raise ValueError(f"{label} must include a timezone.")
+
+
+def _relative_url(value: str) -> None:
+    if (
+        not value.startswith("/")
+        or value.startswith("//")
+        or "\\" in value
+        or len(value) > 1_500
+        or any(ord(character) < 32 or ord(character) == 127 for character in value)
+    ):
+        raise ValueError("Ticket escalation URLs must be bounded application-relative paths.")
+
+
+__all__ = [
+    "CAPABILITY_TICKET_CASE_ESCALATION",
+    "CAPABILITY_TICKET_ROUTING",
+    "TICKET_INTEGRATION_CONTRACT_VERSION",
+    "TicketCaseEscalationCommand",
+    "TicketCaseEscalationProvider",
+    "TicketCaseEscalationResult",
+    "TicketRoutingPlan",
+    "TicketRoutingProvider",
+    "TicketRoutingRequest",
+    "ticket_case_escalation_provider",
+    "ticket_routing_provider",
+]

src/govoplan_core/core/views.py

@@ -2,8 +2,8 @@ from __future__ import annotations
 
 import re
 from collections.abc import Iterable
-from dataclasses import dataclass
-from typing import Literal, Protocol, runtime_checkable
+from dataclasses import dataclass, field
+from typing import Literal, Mapping, Protocol, runtime_checkable
 
 
 VIEWS_MODULE_ID = "views"
@@ -17,6 +17,8 @@ ViewSurfaceKind = Literal[
     "section",
     "action",
     "selector",
+    "product_area",
+    "quick_access",
 ]
 
 _SURFACE_ID_RE = re.compile(r"^[a-z][a-z0-9_.-]{2,159}$")
@@ -42,6 +44,7 @@ class EffectiveView:
     revision_id: str | None
     name: str | None
     visible_surface_ids: frozenset[str]
+    presentation: Mapping[str, object] = field(default_factory=dict)
     locked: bool = False
     projection_active: bool = False
     provenance: tuple[dict[str, object], ...] = ()

src/govoplan_core/db/bootstrap.py

@@ -73,7 +73,7 @@ def bootstrap_dev_data(
 ) -> BootstrapResult:
     tenant = session.query(Tenant).filter(Tenant.slug == tenant_slug).one_or_none()
     if tenant is None:
-        tenant = Tenant(slug=tenant_slug, name="Default Tenant", default_locale="en", settings={})
+        tenant = Tenant(slug=tenant_slug, name="Default Tenant", default_locale="de", settings={})
         session.add(tenant)
         session.flush()
 

src/govoplan_core/db/json_predicates.py

@@ -0,0 +1,147 @@
+"""Exact, bound-parameter JSON permission predicates for SQLite/PostgreSQL.
+
+These primitives only match strings (never coerced numbers/booleans) and
+objects inside actual arrays. Owners still define tenant, subject, permission,
+purpose and current-state policy. Unsupported dialects fail at compilation.
+"""
+from __future__ import annotations
+
+import re
+from collections.abc import Mapping
+
+from sqlalchemy import Boolean, literal
+from sqlalchemy.exc import CompileError
+from sqlalchemy.ext.compiler import compiles
+from sqlalchemy.sql.functions import FunctionElement
+
+
+class _ArrayString(FunctionElement):
+    type = Boolean()
+    inherit_cache = True
+
+
+class _ObjectStrings(FunctionElement):
+    type = Boolean()
+    inherit_cache = True
+
+
+class _ArrayObjectStrings(FunctionElement):
+    type = Boolean()
+    inherit_cache = True
+
+
+def json_array_contains_string(column, value: str):
+    if type(value) is not str:
+        raise TypeError("JSON string membership requires a string value.")
+    return _ArrayString(column, literal(value))
+
+
+def _field_arguments(fields: Mapping[str, str]):
+    if not isinstance(fields, Mapping) or not 1 <= len(fields) <= 16:
+        raise ValueError("JSON object matching requires between 1 and 16 string fields.")
+    arguments = []
+    for key, value in sorted(fields.items()):
+        if type(key) is not str or re.fullmatch(r"[A-Za-z_][A-Za-z0-9_]{0,127}", key) is None:
+            raise ValueError("JSON object field names must be simple identifiers.")
+        if type(value) is not str:
+            raise TypeError("JSON object matching requires string values.")
+        arguments.extend((literal(key), literal(value)))
+    return arguments
+
+
+def json_object_matches_strings(column, fields: Mapping[str, str]):
+    return _ObjectStrings(column, *_field_arguments(fields))
+
+
+def json_array_contains_object_strings(column, fields: Mapping[str, str]):
+    return _ArrayObjectStrings(column, *_field_arguments(fields))
+
+
+@compiles(_ArrayString)
+@compiles(_ObjectStrings)
+@compiles(_ArrayObjectStrings)
+def _unsupported(element, compiler, **kwargs):
+    raise CompileError("Exact JSON permission predicates support only SQLite and PostgreSQL.")
+
+
+def _parts(element, compiler, kwargs):
+    return [compiler.process(item, **kwargs) for item in element.clauses]
+
+
+def _sqlite_array(value):
+    return f"CASE WHEN json_type({value}) = 'array' THEN {value} ELSE '[]' END"
+
+
+def _postgres_array(value):
+    value = f"CAST({value} AS JSON)"
+    return f"CASE WHEN json_typeof({value}) = 'array' THEN {value} ELSE '[]'::json END"
+
+
+def _sqlite_fields(value, fields):
+    terms = []
+    for index in range(0, len(fields), 2):
+        key, expected = fields[index:index + 2]
+        path = f"('$.' || {key})"
+        terms.extend((f"json_type({value}, {path}) = 'text'", f"json_extract({value}, {path}) = {expected}"))
+    return " AND ".join(terms)
+
+
+def _postgres_fields(value, fields):
+    terms = []
+    for index in range(0, len(fields), 2):
+        key, expected = fields[index:index + 2]
+        terms.extend((f"json_typeof(({value}) -> {key}) = 'string'", f"(({value}) ->> {key}) = {expected}"))
+    return " AND ".join(terms)
+
+
+@compiles(_ArrayString, "sqlite")
+def _array_string_sqlite(element, compiler, **kwargs):
+    value, expected = _parts(element, compiler, kwargs)
+    return (
+        f"EXISTS (SELECT 1 FROM json_each({_sqlite_array(value)}) AS gp_json_string "
+        f"WHERE gp_json_string.type = 'text' AND gp_json_string.value = {expected})"
+    )
+
+
+@compiles(_ArrayString, "postgresql")
+def _array_string_postgres(element, compiler, **kwargs):
+    value, expected = _parts(element, compiler, kwargs)
+    return (
+        f"EXISTS (SELECT 1 FROM json_array_elements({_postgres_array(value)}) AS gp_json_string(value) "
+        f"WHERE json_typeof(gp_json_string.value) = 'string' "
+        f"AND (gp_json_string.value #>> '{{}}') = {expected})"
+    )
+
+
+@compiles(_ObjectStrings, "sqlite")
+def _object_sqlite(element, compiler, **kwargs):
+    value, *fields = _parts(element, compiler, kwargs)
+    value = f"CASE WHEN json_type({value}) = 'object' THEN {value} ELSE '{{}}' END"
+    return f"({_sqlite_fields(value, fields)})"
+
+
+@compiles(_ObjectStrings, "postgresql")
+def _object_postgres(element, compiler, **kwargs):
+    value, *fields = _parts(element, compiler, kwargs)
+    value = f"CAST({value} AS JSON)"
+    return f"(json_typeof({value}) = 'object' AND {_postgres_fields(value, fields)})"
+
+
+@compiles(_ArrayObjectStrings, "sqlite")
+def _array_object_sqlite(element, compiler, **kwargs):
+    value, *fields = _parts(element, compiler, kwargs)
+    item = "CASE WHEN gp_json_object.type = 'object' THEN gp_json_object.value ELSE '{}' END"
+    return (
+        f"EXISTS (SELECT 1 FROM json_each({_sqlite_array(value)}) AS gp_json_object "
+        f"WHERE {_sqlite_fields(item, fields)})"
+    )
+
+
+@compiles(_ArrayObjectStrings, "postgresql")
+def _array_object_postgres(element, compiler, **kwargs):
+    value, *fields = _parts(element, compiler, kwargs)
+    return (
+        f"EXISTS (SELECT 1 FROM json_array_elements({_postgres_array(value)}) AS gp_json_object(value) "
+        f"WHERE json_typeof(gp_json_object.value) = 'object' "
+        f"AND {_postgres_fields('gp_json_object.value', fields)})"
+    )

src/govoplan_core/db/migrations.py

@@ -685,7 +685,10 @@ def configured_migration_heads(
         manifest_factories=manifest_factories,
         migration_track=migration_track,
     )
-    return tuple(sorted(ScriptDirectory.from_config(config).get_heads()))
+    scripts = ScriptDirectory.from_config(config)
+    return tuple(
+        sorted(revision.revision for revision in scripts.get_revisions("heads"))
+    )
 
 
 def database_is_at_configured_heads(

src/govoplan_core/db/temporal.py

@@ -0,0 +1,74 @@
+from __future__ import annotations
+
+from typing import Any
+
+from sqlalchemy import or_
+
+from govoplan_core.core.temporal import (
+    TemporalContextError,
+    TemporalDataContext,
+    current_temporal_data_context,
+)
+
+
+def apply_temporal_revision_filter(
+    query: Any,
+    model: type[Any],
+    *,
+    context: TemporalDataContext | None = None,
+    valid_from: str | None = "valid_from",
+    valid_to: str | None = "valid_to",
+    recorded_at: str | None = "recorded_at",
+    superseded_at: str | None = "superseded_at",
+) -> Any:
+    """Apply latest/as-recorded and valid-time clauses to a revision query."""
+
+    resolved = context or current_temporal_data_context()
+    clauses: list[Any] = []
+
+    superseded_column = _optional_column(model, superseded_at)
+    recorded_column = _optional_column(model, recorded_at)
+    if resolved.recorded_at is None:
+        if superseded_column is not None:
+            clauses.append(superseded_column.is_(None))
+    else:
+        if recorded_column is None or superseded_column is None:
+            raise TemporalContextError(
+                f"{model.__name__} does not expose recorded/superseded revision time."
+            )
+        clauses.extend(
+            (
+                recorded_column <= resolved.recorded_at,
+                or_(
+                    superseded_column.is_(None),
+                    superseded_column > resolved.recorded_at,
+                ),
+            )
+        )
+
+    instant = resolved.validity_instant
+    if instant is not None:
+        valid_from_column = _optional_column(model, valid_from)
+        valid_to_column = _optional_column(model, valid_to)
+        if valid_from_column is not None:
+            clauses.append(
+                or_(valid_from_column.is_(None), valid_from_column <= instant)
+            )
+        if valid_to_column is not None:
+            clauses.append(or_(valid_to_column.is_(None), valid_to_column > instant))
+
+    return query.filter(*clauses) if clauses else query
+
+
+def _optional_column(model: type[Any], name: str | None) -> Any | None:
+    if name is None:
+        return None
+    column = getattr(model, name, None)
+    if column is None:
+        raise TemporalContextError(
+            f"{model.__name__} has no temporal column named {name!r}."
+        )
+    return column
+
+
+__all__ = ["apply_temporal_revision_filter"]

src/govoplan_core/i18n.py

@@ -4,10 +4,12 @@ import re
 from typing import Any, Iterable
 
 I18N_SETTINGS_KEY = "i18n"
+REFERENCE_LANGUAGE_CODE = "de"
+SOURCE_LANGUAGE_CODE = "en"
 
 DEFAULT_LANGUAGE_PACKAGES: tuple[dict[str, str], ...] = (
-    {"code": "en", "label": "English", "native_label": "English"},
     {"code": "de", "label": "German", "native_label": "Deutsch"},
+    {"code": "en", "label": "English", "native_label": "English"},
 )
 
 
@@ -66,7 +68,7 @@ def normalize_enabled_language_codes(
     available_languages: Iterable[dict[str, Any]],
     *,
     default_locale: object = None,
-    fallback_codes: Iterable[str] = ("en", "de"),
+    fallback_codes: Iterable[str] = (REFERENCE_LANGUAGE_CODE, SOURCE_LANGUAGE_CODE),
 ) -> list[str]:
     available_codes = [normalize_language_code(item.get("code")) for item in available_languages if isinstance(item, dict)]
     available = {code for code in available_codes if code}
@@ -88,8 +90,10 @@ def normalize_enabled_language_codes(
 
     if enabled:
         return enabled
-    if "en" in available:
-        return ["en"]
+    if REFERENCE_LANGUAGE_CODE in available:
+        return [REFERENCE_LANGUAGE_CODE]
+    if SOURCE_LANGUAGE_CODE in available:
+        return [SOURCE_LANGUAGE_CODE]
     return available_codes[:1]
 
 
@@ -126,7 +130,7 @@ def preferred_language_code(user_settings: dict[str, Any] | None, allowed_codes:
     default = resolve_language_code(default_locale, enabled)
     if default:
         return default
-    return enabled[0] if enabled else "en"
+    return enabled[0] if enabled else REFERENCE_LANGUAGE_CODE
 
 
 def update_i18n_settings(settings: dict[str, Any] | None, **values: Any) -> dict[str, Any]:
@@ -141,10 +145,13 @@ def system_i18n_payload(settings_item: Any | None) -> dict[str, object]:
     settings = getattr(settings_item, "settings", None)
     packages = system_language_packages(settings)
     available_codes = [item["code"] for item in packages]
-    default_language = resolve_language_code(getattr(settings_item, "default_locale", None), available_codes) or "en"
+    default_language = (
+        resolve_language_code(getattr(settings_item, "default_locale", None), available_codes)
+        or REFERENCE_LANGUAGE_CODE
+    )
     enabled = system_enabled_language_codes(settings, default_locale=default_language)
     if default_language not in enabled:
-        default_language = enabled[0] if enabled else "en"
+        default_language = enabled[0] if enabled else REFERENCE_LANGUAGE_CODE
     return {
         "available_languages": packages,
         "enabled_languages": enabled,

src/govoplan_core/mail/__init__.py

@@ -1,3 +1,8 @@
-from govoplan_core.mail.config import ImapConfig, SmtpConfig, TransportSecurity
+from govoplan_core.mail.config import (
+    ImapConfig,
+    ImapFolderMappings,
+    SmtpConfig,
+    TransportSecurity,
+)
 
-__all__ = ["ImapConfig", "SmtpConfig", "TransportSecurity"]
+__all__ = ["ImapConfig", "ImapFolderMappings", "SmtpConfig", "TransportSecurity"]

src/govoplan_core/mail/config.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from enum import StrEnum
 from typing import Any
 
-from pydantic import BaseModel, ConfigDict, Field, model_validator
+from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
 
 
 class StrictModel(BaseModel):
@@ -34,21 +34,61 @@ class SmtpServerConfig(StrictModel):
         return self
 
 
+class ImapFolderMappings(StrictModel):
+    """Profile-level names for the standard IMAP mailbox roles."""
+
+    inbox: str | None = None
+    sent: str | None = None
+    drafts: str | None = None
+    trash: str | None = None
+    archive: str | None = None
+    junk: str | None = None
+
+    @field_validator("*", mode="before")
+    @classmethod
+    def normalize_folder_name(cls, value: Any) -> Any:
+        if value is None:
+            return None
+        normalized = str(value).strip()
+        return normalized or None
+
+
 class ImapServerConfig(StrictModel):
     host: str | None = None
     port: int | None = Field(default=None, ge=1, le=65535)
     security: TransportSecurity = TransportSecurity.TLS
     sent_folder: str = "auto"
+    folder_mappings: ImapFolderMappings | None = None
     timeout_seconds: int = Field(default=30, ge=1)
 
     @model_validator(mode="before")
     @classmethod
     def discard_legacy_enabled(cls, value: Any) -> Any:
-        if isinstance(value, dict) and "enabled" in value:
+        if not isinstance(value, dict):
+            return value
         data = dict(value)
         data.pop("enabled", None)
+        mappings_value = data.get("folder_mappings")
+        mappings = (
+            mappings_value.model_dump(exclude_none=True)
+            if isinstance(mappings_value, ImapFolderMappings)
+            else dict(mappings_value)
+            if isinstance(mappings_value, dict)
+            else {}
+        )
+        mapped_sent = str(mappings.get("sent") or "").strip()
+        legacy_sent = str(data.get("sent_folder") or "").strip()
+        if mapped_sent:
+            # The typed mapping is canonical when both new and legacy callers
+            # provide a Sent value. Keep the legacy field synchronized for
+            # existing Campaign append consumers.
+            data["sent_folder"] = mapped_sent
+        elif legacy_sent and legacy_sent != "auto":
+            data["sent_folder"] = legacy_sent
+            mappings["sent"] = legacy_sent
+        if mappings:
+            data["folder_mappings"] = mappings
         return data
-        return value
 
     @model_validator(mode="after")
     def apply_default_port(self) -> "ImapServerConfig":

src/govoplan_core/privacy/dsar_workflow.py

@@ -0,0 +1,753 @@
+from __future__ import annotations
+
+import hashlib
+import json
+import uuid
+from collections import defaultdict
+from collections.abc import Iterable, Mapping, Sequence
+from datetime import datetime, timezone
+from typing import Any
+
+from sqlalchemy import DateTime, Index, Integer, JSON, String, Text
+from sqlalchemy.orm import Mapped, Session, mapped_column
+
+from govoplan_core.core.concurrency import RevisionConflictError, strong_resource_etag
+from govoplan_core.core.dsar import (
+    DsarErasureActionRef,
+    DsarExecutionResultRef,
+    DsarProvider,
+    DsarRecordRef,
+    DsarRequestKind,
+    DsarSubjectRef,
+    dsar_provider_names,
+)
+from govoplan_core.db.base import Base, TimestampMixin
+
+
+MAX_RECORDS_PER_PROVIDER = 10_000
+MAX_PROVIDER_RESULT_BYTES = 10 * 1024 * 1024
+
+
+class DataSubjectRequest(Base, TimestampMixin):
+    __tablename__ = "core_data_subject_requests"
+    __table_args__ = (
+        Index(
+            "ix_core_data_subject_requests_tenant_status",
+            "tenant_id",
+            "status",
+        ),
+    )
+
+    id: Mapped[str] = mapped_column(
+        String(36), primary_key=True, default=lambda: str(uuid.uuid4())
+    )
+    tenant_id: Mapped[str] = mapped_column(String(36), nullable=False, index=True)
+    reference: Mapped[str] = mapped_column(String(120), nullable=False)
+    request_kind: Mapped[str] = mapped_column(String(30), nullable=False)
+    status: Mapped[str] = mapped_column(
+        String(30), default="draft", nullable=False, index=True
+    )
+    subject: Mapped[dict[str, Any]] = mapped_column(JSON, default=dict, nullable=False)
+    purpose: Mapped[str] = mapped_column(String(1000), nullable=False)
+    legal_basis: Mapped[str | None] = mapped_column(String(1000), nullable=True)
+    due_at: Mapped[datetime | None] = mapped_column(
+        DateTime(timezone=True), nullable=True, index=True
+    )
+    requested_by_account_id: Mapped[str] = mapped_column(String(36), nullable=False)
+    search_result: Mapped[dict[str, Any]] = mapped_column(
+        JSON, default=dict, nullable=False
+    )
+    erasure_plan: Mapped[dict[str, Any]] = mapped_column(
+        JSON, default=dict, nullable=False
+    )
+    execution_result: Mapped[dict[str, Any]] = mapped_column(
+        JSON, default=dict, nullable=False
+    )
+    coverage: Mapped[dict[str, Any]] = mapped_column(JSON, default=dict, nullable=False)
+    evidence_sha256: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    resource_revision: Mapped[int] = mapped_column(Integer, default=1, nullable=False)
+    completed_at: Mapped[datetime | None] = mapped_column(
+        DateTime(timezone=True), nullable=True
+    )
+    notes: Mapped[str | None] = mapped_column(Text, nullable=True)
+
+    @property
+    def strong_etag(self) -> str:
+        return strong_resource_etag(
+            "data_subject_request",
+            self.id,
+            self.resource_revision,
+        )
+
+
+def create_data_subject_request(
+    session: Session,
+    *,
+    tenant_id: str,
+    reference: str,
+    request_kind: DsarRequestKind,
+    subject: DsarSubjectRef,
+    purpose: str,
+    legal_basis: str | None,
+    due_at: datetime | None,
+    requested_by_account_id: str,
+    notes: str | None = None,
+) -> DataSubjectRequest:
+    if not subject.has_selector():
+        raise ValueError("At least one data-subject selector is required.")
+    row = DataSubjectRequest(
+        tenant_id=tenant_id,
+        reference=reference.strip(),
+        request_kind=request_kind,
+        subject=subject.to_dict(),
+        purpose=purpose.strip(),
+        legal_basis=(legal_basis or "").strip() or None,
+        due_at=due_at,
+        requested_by_account_id=requested_by_account_id,
+        notes=(notes or "").strip() or None,
+    )
+    session.add(row)
+    session.flush()
+    row.evidence_sha256 = _evidence_digest(row)
+    return row
+
+
+def get_data_subject_request(
+    session: Session,
+    *,
+    tenant_id: str,
+    request_id: str,
+    for_update: bool = False,
+) -> DataSubjectRequest:
+    query = session.query(DataSubjectRequest).filter(
+        DataSubjectRequest.id == request_id,
+        DataSubjectRequest.tenant_id == tenant_id,
+    )
+    if for_update:
+        query = query.with_for_update()
+    row = query.one_or_none()
+    if row is None or row.tenant_id != tenant_id:
+        raise LookupError("Data-subject request not found.")
+    return row
+
+
+def list_data_subject_requests(
+    session: Session,
+    *,
+    tenant_id: str,
+    limit: int = 200,
+) -> tuple[DataSubjectRequest, ...]:
+    return tuple(
+        session.query(DataSubjectRequest)
+        .filter(DataSubjectRequest.tenant_id == tenant_id)
+        .order_by(
+            DataSubjectRequest.created_at.desc(),
+            DataSubjectRequest.id.desc(),
+        )
+        .limit(max(1, min(limit, 500)))
+        .all()
+    )
+
+
+def search_data_subject_request(
+    session: Session,
+    *,
+    registry: object,
+    row: DataSubjectRequest,
+    expected_revision: int,
+) -> DataSubjectRequest:
+    _assert_revision(row, expected_revision)
+    subject = _subject(row.subject)
+    records: list[dict[str, object]] = []
+    provider_runs: list[dict[str, object]] = []
+    providers, discovery = _providers(
+        registry,
+        session,
+        tenant_id=row.tenant_id,
+    )
+    for capability_name, provider in providers:
+        started_at = _now()
+        try:
+            with session.begin_nested():
+                provider_records = tuple(
+                    provider.search_subject(
+                        session,
+                        tenant_id=row.tenant_id,
+                        subject=subject,
+                    )
+                )
+                _validate_records(provider, provider_records)
+                if len(provider_records) > MAX_RECORDS_PER_PROVIDER:
+                    raise ValueError("DSAR provider result exceeds the record limit.")
+                encoded = _json_bytes([item.to_dict() for item in provider_records])
+                if len(encoded) > MAX_PROVIDER_RESULT_BYTES:
+                    raise ValueError("DSAR provider result exceeds the payload limit.")
+            records.extend(item.to_dict() for item in provider_records)
+            provider_runs.append(
+                {
+                    "capability": capability_name,
+                    "provider_id": provider.provider_id,
+                    "module_id": provider.module_id,
+                    "status": "complete",
+                    "record_count": len(provider_records),
+                    "started_at": started_at,
+                    "completed_at": _now(),
+                }
+            )
+        except Exception as exc:
+            provider_runs.append(
+                {
+                    "capability": capability_name,
+                    "provider_id": getattr(provider, "provider_id", capability_name),
+                    "module_id": getattr(provider, "module_id", "unknown"),
+                    "status": "failed",
+                    "record_count": 0,
+                    "error": _safe_error(exc),
+                    "started_at": started_at,
+                    "completed_at": _now(),
+                }
+            )
+    failures = sum(1 for item in provider_runs if item["status"] == "failed")
+    row.search_result = {
+        "schema": "govoplan.dsars.search.v1",
+        "searched_at": _now(),
+        "records": records,
+        "provider_runs": provider_runs,
+        "record_count": len(records),
+    }
+    row.coverage = discovery
+    row.erasure_plan = {}
+    row.execution_result = {}
+    row.status = "search_partial" if failures else "searched"
+    row.completed_at = None
+    _advance(row)
+    return row
+
+
+def plan_data_subject_erasure(
+    session: Session,
+    *,
+    registry: object,
+    row: DataSubjectRequest,
+    expected_revision: int,
+) -> DataSubjectRequest:
+    _assert_revision(row, expected_revision)
+    if not row.search_result.get("searched_at"):
+        raise ValueError("Run the data-subject search before planning erasure.")
+    if row.request_kind == "access":
+        raise ValueError("This request does not include erasure.")
+    subject = _subject(row.subject)
+    records = tuple(_record(item) for item in row.search_result.get("records", []))
+    by_provider: dict[str, list[DsarRecordRef]] = defaultdict(list)
+    for record in records:
+        by_provider[record.provider_id].append(record)
+    providers, discovery = _providers(
+        registry,
+        session,
+        tenant_id=row.tenant_id,
+    )
+    actions: list[DsarErasureActionRef] = []
+    provider_runs: list[dict[str, object]] = []
+    for capability_name, provider in providers:
+        started_at = _now()
+        try:
+            with session.begin_nested():
+                proposed = tuple(
+                    provider.plan_erasure(
+                        session,
+                        tenant_id=row.tenant_id,
+                        subject=subject,
+                        records=tuple(by_provider.get(provider.provider_id, ())),
+                    )
+                )
+                _validate_actions(provider, proposed)
+            actions.extend(proposed)
+            provider_runs.append(
+                {
+                    "capability": capability_name,
+                    "provider_id": provider.provider_id,
+                    "module_id": provider.module_id,
+                    "status": "complete",
+                    "action_count": len(proposed),
+                    "started_at": started_at,
+                    "completed_at": _now(),
+                }
+            )
+        except Exception as exc:
+            provider_runs.append(
+                {
+                    "capability": capability_name,
+                    "provider_id": getattr(provider, "provider_id", capability_name),
+                    "module_id": getattr(provider, "module_id", "unknown"),
+                    "status": "failed",
+                    "action_count": 0,
+                    "error": _safe_error(exc),
+                    "started_at": started_at,
+                    "completed_at": _now(),
+                }
+            )
+    existing_resources = {
+        (item.provider_id, item.resource_type, item.resource_id) for item in actions
+    }
+    for record in records:
+        key = (record.provider_id, record.resource_type, record.resource_id)
+        if not record.immutable_evidence or key in existing_resources:
+            continue
+        digest = hashlib.sha256("\0".join(key).encode("utf-8")).hexdigest()[:24]
+        actions.append(
+            DsarErasureActionRef(
+                action_id=f"retain:{digest}",
+                provider_id=record.provider_id,
+                module_id=record.module_id,
+                kind="retain",
+                resource_type=record.resource_type,
+                resource_id=record.resource_id,
+                title=f"Retain {record.title}",
+                rationale=(
+                    record.retention_reason
+                    or "Immutable institutional evidence must be retained."
+                ),
+                executable=False,
+            )
+        )
+    duplicate_ids = _duplicates(item.action_id for item in actions)
+    if duplicate_ids:
+        raise ValueError(
+            f"DSAR providers returned duplicate action ids: {', '.join(duplicate_ids)}"
+        )
+    failures = sum(1 for item in provider_runs if item["status"] == "failed")
+    row.erasure_plan = {
+        "schema": "govoplan.dsars.erasure-plan.v1",
+        "planned_at": _now(),
+        "actions": [item.to_dict() for item in actions],
+        "provider_runs": provider_runs,
+        "executable_count": sum(1 for item in actions if item.executable),
+        "retained_count": sum(1 for item in actions if item.kind == "retain"),
+    }
+    row.coverage = {**discovery, "plan_provider_runs": provider_runs}
+    row.execution_result = {}
+    row.status = "plan_partial" if failures else "plan_ready"
+    row.completed_at = None
+    _advance(row)
+    return row
+
+
+def execute_data_subject_erasure(
+    session: Session,
+    *,
+    registry: object,
+    row: DataSubjectRequest,
+    expected_revision: int,
+    action_ids: Sequence[str],
+) -> DataSubjectRequest:
+    _assert_revision(row, expected_revision)
+    raw_actions = row.erasure_plan.get("actions")
+    if not isinstance(raw_actions, list):
+        raise ValueError("Create an erasure plan before execution.")
+    actions_by_id = {
+        action.action_id: action for action in (_action(item) for item in raw_actions)
+    }
+    selected_ids = tuple(dict.fromkeys(str(item) for item in action_ids if str(item)))
+    if not selected_ids:
+        raise ValueError("Select at least one executable erasure action.")
+    missing = [item for item in selected_ids if item not in actions_by_id]
+    if missing:
+        raise ValueError("The erasure plan changed; refresh before executing it.")
+    selected = tuple(actions_by_id[item] for item in selected_ids)
+    blocked = [item.action_id for item in selected if not item.executable]
+    if blocked:
+        raise ValueError("Retained or review-only actions cannot be executed.")
+
+    providers, _discovery = _providers(
+        registry,
+        session,
+        tenant_id=row.tenant_id,
+    )
+    providers_by_id = {provider.provider_id: provider for _name, provider in providers}
+    grouped: dict[str, list[DsarErasureActionRef]] = defaultdict(list)
+    for action in selected:
+        grouped[action.provider_id].append(action)
+    previous = {
+        str(item.get("action_id")): item
+        for item in row.execution_result.get("results", [])
+        if isinstance(item, Mapping)
+        and item.get("status") in {"executed", "unchanged"}
+    }
+    results: list[dict[str, object]] = list(previous.values())
+    for provider_id, provider_actions in grouped.items():
+        provider = providers_by_id.get(provider_id)
+        if provider is None:
+            results.extend(
+                DsarExecutionResultRef(
+                    action_id=action.action_id,
+                    status="blocked",
+                    summary="The owning DSAR provider is not currently available.",
+                ).to_dict()
+                for action in provider_actions
+            )
+            continue
+        pending = tuple(
+            action for action in provider_actions if action.action_id not in previous
+        )
+        if not pending:
+            continue
+        try:
+            with session.begin_nested():
+                executed = tuple(
+                    provider.execute_erasure(
+                        session,
+                        tenant_id=row.tenant_id,
+                        subject=_subject(row.subject),
+                        actions=pending,
+                        request_id=row.id,
+                    )
+                )
+                _validate_execution_results(pending, executed)
+            results.extend(item.to_dict() for item in executed)
+        except Exception as exc:
+            results.extend(
+                DsarExecutionResultRef(
+                    action_id=action.action_id,
+                    status="failed",
+                    summary=_safe_error(exc),
+                ).to_dict()
+                for action in pending
+            )
+    result_by_id = {str(item["action_id"]): item for item in results}
+    all_executable = {
+        action.action_id for action in actions_by_id.values() if action.executable
+    }
+    successful = {
+        action_id
+        for action_id, item in result_by_id.items()
+        if item.get("status") in {"executed", "unchanged"}
+    }
+    failed = {
+        action_id
+        for action_id, item in result_by_id.items()
+        if item.get("status") in {"failed", "blocked"}
+    }
+    row.execution_result = {
+        "schema": "govoplan.dsars.execution.v1",
+        "executed_at": _now(),
+        "results": list(result_by_id.values()),
+        "successful_count": len(successful),
+        "failed_count": len(failed),
+    }
+    if all_executable.issubset(successful):
+        row.status = "completed"
+        row.completed_at = datetime.now(timezone.utc)
+    elif failed:
+        row.status = "execution_partial"
+        row.completed_at = None
+    else:
+        row.status = "execution_pending"
+        row.completed_at = None
+    _advance(row)
+    return row
+
+
+def data_subject_export(row: DataSubjectRequest) -> bytes:
+    payload: dict[str, object] = {
+        "schema": "govoplan.dsars.export.v1",
+        "generated_at": _now(),
+        "request": data_subject_request_dict(row, include_subject=True),
+        "search": row.search_result,
+        "erasure_plan": row.erasure_plan,
+        "execution": row.execution_result,
+        "coverage": row.coverage,
+        "limitations": [
+            "Only providers listed as complete contributed data.",
+            "Retained immutable evidence is exported with its stated retention reason.",
+            "An export does not itself erase or alter source data.",
+        ],
+    }
+    digest = hashlib.sha256(_json_bytes(payload)).hexdigest()
+    payload["manifest_sha256"] = digest
+    return _json_bytes(payload, pretty=True)
+
+
+def data_subject_request_dict(
+    row: DataSubjectRequest,
+    *,
+    include_subject: bool,
+) -> dict[str, object]:
+    return {
+        "id": row.id,
+        "tenant_id": row.tenant_id,
+        "reference": row.reference,
+        "request_kind": row.request_kind,
+        "status": row.status,
+        "subject": dict(row.subject) if include_subject else {},
+        "purpose": row.purpose,
+        "legal_basis": row.legal_basis,
+        "due_at": row.due_at.isoformat() if row.due_at else None,
+        "requested_by_account_id": row.requested_by_account_id,
+        "record_count": int(row.search_result.get("record_count", 0)),
+        "executable_action_count": int(
+            row.erasure_plan.get("executable_count", 0)
+        ),
+        "coverage": dict(row.coverage),
+        "evidence_sha256": row.evidence_sha256,
+        "resource_revision": row.resource_revision,
+        "etag": row.strong_etag,
+        "created_at": row.created_at.isoformat(),
+        "updated_at": row.updated_at.isoformat(),
+        "completed_at": row.completed_at.isoformat() if row.completed_at else None,
+        "notes": row.notes,
+    }
+
+
+def _providers(
+    registry: object,
+    session: Session,
+    *,
+    tenant_id: str,
+) -> tuple[tuple[tuple[str, DsarProvider], ...], dict[str, object]]:
+    values: list[tuple[str, DsarProvider]] = []
+    failures: list[dict[str, str]] = []
+    active_modules = _effective_module_ids(registry, session, tenant_id=tenant_id)
+    names = dsar_provider_names(registry)
+    active_names: list[str] = []
+    inactive_names: list[str] = []
+    for name in names:
+        owner = (
+            registry.capability_owner(name)
+            if hasattr(registry, "capability_owner")
+            else None
+        )
+        if owner and owner not in active_modules:
+            inactive_names.append(name)
+            continue
+        active_names.append(name)
+        try:
+            if hasattr(registry, "require_tenant_capability"):
+                candidate = registry.require_tenant_capability(
+                    name,
+                    session,
+                    tenant_id=tenant_id,
+                    work_state="interactive",
+                )
+            else:
+                candidate = registry.require_capability(name)
+            if not isinstance(candidate, DsarProvider):
+                raise TypeError("Capability does not implement DsarProvider.")
+            values.append((name, candidate))
+        except Exception as exc:
+            failures.append({"capability": name, "error": _safe_error(exc)})
+    covered_modules = sorted({provider.module_id for _name, provider in values})
+    return tuple(values), {
+        "provider_capabilities": active_names,
+        "inactive_provider_capabilities": inactive_names,
+        "covered_modules": covered_modules,
+        "modules_without_provider": sorted(set(active_modules) - set(covered_modules)),
+        "provider_discovery_failures": failures,
+    }
+
+
+def _effective_module_ids(
+    registry: object,
+    session: Session,
+    *,
+    tenant_id: str,
+) -> list[str]:
+    resolver_factory = getattr(registry, "tenant_entitlement_resolver", None)
+    if callable(resolver_factory):
+        state = resolver_factory().resolve(session, tenant_id)
+        return sorted(str(item) for item in state.effective_modules)
+    if hasattr(registry, "manifests"):
+        return sorted(str(manifest.id) for manifest in registry.manifests())
+    return []
+
+
+def _subject(value: Mapping[str, object]) -> DsarSubjectRef:
+    raw_refs = value.get("external_references")
+    return DsarSubjectRef(
+        account_id=_optional_text(value.get("account_id")),
+        identity_id=_optional_text(value.get("identity_id")),
+        membership_id=_optional_text(value.get("membership_id")),
+        email=_optional_text(value.get("email")),
+        external_references={
+            str(key): str(item)
+            for key, item in (raw_refs.items() if isinstance(raw_refs, Mapping) else ())
+        },
+    )
+
+
+def _record(value: object) -> DsarRecordRef:
+    if not isinstance(value, Mapping):
+        raise ValueError("Stored DSAR record is invalid.")
+    observed_at = _parse_datetime(value.get("observed_at"))
+    data = value.get("data")
+    return DsarRecordRef(
+        provider_id=str(value.get("provider_id") or ""),
+        module_id=str(value.get("module_id") or ""),
+        resource_type=str(value.get("resource_type") or ""),
+        resource_id=str(value.get("resource_id") or ""),
+        category=str(value.get("category") or ""),
+        title=str(value.get("title") or ""),
+        data=dict(data) if isinstance(data, Mapping) else {},
+        observed_at=observed_at,
+        immutable_evidence=bool(value.get("immutable_evidence")),
+        retention_reason=_optional_text(value.get("retention_reason")),
+        source_path=_optional_text(value.get("source_path")),
+    )
+
+
+def _action(value: object) -> DsarErasureActionRef:
+    if not isinstance(value, Mapping):
+        raise ValueError("Stored DSAR action is invalid.")
+    metadata = value.get("metadata")
+    kind = str(value.get("kind") or "manual_review")
+    if kind not in {
+        "delete",
+        "anonymize",
+        "revoke",
+        "detach",
+        "retain",
+        "manual_review",
+    }:
+        raise ValueError("Stored DSAR action kind is invalid.")
+    return DsarErasureActionRef(
+        action_id=str(value.get("action_id") or ""),
+        provider_id=str(value.get("provider_id") or ""),
+        module_id=str(value.get("module_id") or ""),
+        kind=kind,  # type: ignore[arg-type]
+        resource_type=str(value.get("resource_type") or ""),
+        resource_id=str(value.get("resource_id") or ""),
+        title=str(value.get("title") or ""),
+        rationale=str(value.get("rationale") or ""),
+        executable=bool(value.get("executable")),
+        irreversible=bool(value.get("irreversible")),
+        metadata=dict(metadata) if isinstance(metadata, Mapping) else {},
+    )
+
+
+def _validate_records(
+    provider: DsarProvider,
+    records: Sequence[DsarRecordRef],
+) -> None:
+    for record in records:
+        if record.provider_id != provider.provider_id or record.module_id != provider.module_id:
+            raise ValueError("DSAR record ownership does not match its provider.")
+        if not record.resource_type or not record.resource_id or not record.title:
+            raise ValueError("DSAR records require resource identity and title.")
+
+
+def _validate_actions(
+    provider: DsarProvider,
+    actions: Sequence[DsarErasureActionRef],
+) -> None:
+    for action in actions:
+        if action.provider_id != provider.provider_id or action.module_id != provider.module_id:
+            raise ValueError("DSAR action ownership does not match its provider.")
+        if not action.action_id or not action.resource_type or not action.resource_id:
+            raise ValueError("DSAR actions require stable identities.")
+        if action.kind in {"retain", "manual_review"} and action.executable:
+            raise ValueError("Retain and manual-review actions cannot be executable.")
+
+
+def _validate_execution_results(
+    actions: Sequence[DsarErasureActionRef],
+    results: Sequence[DsarExecutionResultRef],
+) -> None:
+    expected = {action.action_id for action in actions}
+    returned = {result.action_id for result in results}
+    if expected != returned or len(returned) != len(results):
+        raise ValueError("DSAR provider did not return one result per action.")
+
+
+def _assert_revision(row: DataSubjectRequest, expected_revision: int) -> None:
+    if row.resource_revision != expected_revision:
+        raise RevisionConflictError(
+            resource_type="data_subject_request",
+            resource_id=row.id,
+            current_revision=row.resource_revision,
+            submitted_base_revision=expected_revision,
+            current_etag=row.strong_etag,
+            refresh_path=f"/api/v1/admin/privacy/data-subject-requests/{row.id}",
+        )
+
+
+def _advance(row: DataSubjectRequest) -> None:
+    row.resource_revision += 1
+    row.evidence_sha256 = _evidence_digest(row)
+
+
+def _evidence_digest(row: DataSubjectRequest) -> str:
+    return hashlib.sha256(
+        _json_bytes(
+            {
+                "id": row.id,
+                "tenant_id": row.tenant_id,
+                "reference": row.reference,
+                "request_kind": row.request_kind,
+                "status": row.status,
+                "subject": row.subject,
+                "search_result": row.search_result,
+                "erasure_plan": row.erasure_plan,
+                "execution_result": row.execution_result,
+                "coverage": row.coverage,
+                "resource_revision": row.resource_revision,
+            }
+        )
+    ).hexdigest()
+
+
+def _duplicates(values: Iterable[str]) -> tuple[str, ...]:
+    seen: set[str] = set()
+    duplicate: set[str] = set()
+    for value in values:
+        if value in seen:
+            duplicate.add(value)
+        seen.add(value)
+    return tuple(sorted(duplicate))
+
+
+def _json_bytes(value: object, *, pretty: bool = False) -> bytes:
+    return json.dumps(
+        value,
+        ensure_ascii=False,
+        sort_keys=True,
+        indent=2 if pretty else None,
+        separators=None if pretty else (",", ":"),
+        default=str,
+    ).encode("utf-8")
+
+
+def _safe_error(exc: Exception) -> str:
+    text = " ".join(str(exc).split())
+    return (text or exc.__class__.__name__)[:1000]
+
+
+def _optional_text(value: object | None) -> str | None:
+    text = str(value or "").strip()
+    return text or None
+
+
+def _parse_datetime(value: object | None) -> datetime | None:
+    if isinstance(value, datetime):
+        return value
+    if isinstance(value, str) and value:
+        try:
+            return datetime.fromisoformat(value)
+        except ValueError:
+            return None
+    return None
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+__all__ = [
+    "DataSubjectRequest",
+    "create_data_subject_request",
+    "data_subject_export",
+    "data_subject_request_dict",
+    "execute_data_subject_erasure",
+    "get_data_subject_request",
+    "list_data_subject_requests",
+    "plan_data_subject_erasure",
+    "search_data_subject_request",
+]

src/govoplan_core/resources/catalog-keyring.json

@@ -0,0 +1,13 @@
+{
+  "generated_at": "2026-07-11T15:18:35.649400Z",
+  "keyring_version": "1",
+  "keys": [
+    {
+      "key_id": "release-key-1",
+      "not_before": "2026-07-11T00:00:00Z",
+      "public_key": "jOXIlZXytoNJCH8tsmrYRklg6ShpjGXRY0uV3jApRiA=",
+      "status": "active"
+    }
+  ],
+  "purpose": "govoplan module package catalog signatures"
+}

src/govoplan_core/security/bounded_process.py

@@ -0,0 +1,285 @@
+"""Resource-bounded, disposable workers for trusted module-owned byte operations.
+
+This is not an arbitrary-code sandbox. Callers supply a server-owned top-level
+function, never a client-selected module/path/callable. Sessions, credentials and
+authority remain in the parent; only explicit bounded bytes cross the pipe.
+"""
+from __future__ import annotations
+
+from collections.abc import Callable
+from contextlib import contextmanager
+from dataclasses import dataclass
+import inspect
+import math
+import os
+from pathlib import Path
+import selectors
+import signal
+import subprocess
+import sys
+import threading
+import time
+
+
+@dataclass(frozen=True, slots=True)
+class ProcessLimits:
+    wall_seconds: float = 10.0
+    cpu_seconds: int = 10
+    memory_bytes: int = 256 * 1024 * 1024
+    input_bytes: int = 8 * 1024 * 1024
+    output_bytes: int = 8 * 1024 * 1024
+    file_bytes: int = 0
+
+    def __post_init__(self) -> None:
+        if not math.isfinite(self.wall_seconds) or not 0 < self.wall_seconds <= 600:
+            raise ValueError("Worker wall time must be finite and within (0, 600] seconds.")
+        for name, minimum, maximum in (
+            ("cpu_seconds", 1, 600),
+            ("memory_bytes", 64 * 1024 * 1024, 8 * 1024 * 1024 * 1024),
+            ("input_bytes", 1, 256 * 1024 * 1024),
+            ("output_bytes", 1, 256 * 1024 * 1024),
+            ("file_bytes", 0, 2 * 1024 * 1024 * 1024),
+        ):
+            value = getattr(self, name)
+            if type(value) is not int or not minimum <= value <= maximum:
+                raise ValueError(f"Invalid worker {name} limit.")
+
+
+class ProcessBudgetError(RuntimeError):
+    def __init__(self, code: str) -> None:
+        messages = {
+            "busy": "The isolated-work capacity is busy; retry later.",
+            "cancelled": "Isolated work was cancelled.",
+            "timeout": "Isolated work exceeded its wall-clock limit.",
+            "cpu_limit": "Isolated work exceeded its CPU limit.",
+            "memory_limit": "Isolated work exceeded its memory limit.",
+            "input_limit": "Isolated work input exceeded its byte limit.",
+            "output_limit": "Isolated work output exceeded its byte limit.",
+            "unavailable": "Required isolated-worker resource controls are unavailable.",
+            "worker_failed": "Isolated work could not complete safely.",
+        }
+        self.code = code
+        super().__init__(messages[code])
+
+
+_gate = threading.Lock()
+_active = 0
+_STDERR_LIMIT = 64 * 1024
+
+
+def _reset_after_fork() -> None:
+    global _gate, _active
+    _gate, _active = threading.Lock(), 0
+
+
+if hasattr(os, "register_at_fork"):
+    os.register_at_fork(after_in_child=_reset_after_fork)
+
+
+def _reserve() -> None:
+    from govoplan_core.settings import settings
+
+    global _active
+    with _gate:
+        if _active >= settings.isolated_process_concurrency:
+            raise ProcessBudgetError("busy")
+        _active += 1
+
+
+def _release() -> None:
+    global _active
+    with _gate:
+        _active -= 1
+
+
+@dataclass(slots=True, eq=False)
+class _OperationAdmission:
+    process_id: int
+    thread_id: int
+    active: bool = True
+    running: bool = False
+
+
+@contextmanager
+def bounded_operation_admission():
+    """Reserve shared capacity before bounded parent-side input preparation.
+
+    The yielded token may be reused sequentially in this thread only. Never
+    expose it to clients or hold it while waiting for user input/network work.
+    Parent preparation still needs its own byte/time/disk bounds.
+    """
+    _reserve()
+    admission = _OperationAdmission(os.getpid(), threading.get_ident())
+    try:
+        yield admission
+    finally:
+        admission.active = False
+        # Fork resets the child's admission counter. An inherited context must
+        # still expire its token, but cannot release the parent's reservation.
+        if admission.process_id == os.getpid():
+            _release()
+
+
+def _enter_admission(admission: _OperationAdmission) -> None:
+    with _gate:
+        if (
+            not isinstance(admission, _OperationAdmission)
+            or not admission.active or admission.running
+            or admission.process_id != os.getpid()
+            or admission.thread_id != threading.get_ident()
+        ):
+            raise ValueError("Worker admission must be active, unused and owned by this thread/process.")
+        admission.running = True
+
+
+def run_bounded_operation(
+    operation: Callable[[bytes], bytes],
+    payload: bytes,
+    *,
+    limits: ProcessLimits = ProcessLimits(),
+    cancelled: Callable[[], bool] | None = None,
+    admission: _OperationAdmission | None = None,
+) -> bytes:
+    """Run a trusted, importable module function without inheriting parent state.
+
+    Admission is non-queuing and per API/worker process. POSIX process groups,
+    resource limits and waitid(WNOWAIT) are required; no in-process fallback.
+    Output and stderr are drained incrementally, including while input is sent.
+    Every exit path kills the owned process group before reaping its leader.
+    """
+    if type(payload) is not bytes or len(payload) > limits.input_bytes:
+        raise ProcessBudgetError("input_limit")
+    if os.name != "posix" or not hasattr(os, "WNOWAIT") or not hasattr(os, "waitid"):
+        raise ProcessBudgetError("unavailable")
+    module = inspect.getmodule(operation)
+    name = getattr(operation, "__name__", "")
+    if (
+        module is None or not name.isidentifier()
+        or getattr(module, name, None) is not operation
+        or not inspect.isfunction(operation) or not getattr(module, "__file__", None)
+    ):
+        raise ValueError("Isolated operations must be server-owned module-level functions.")
+    module_name = module.__name__
+    if not all(part.isidentifier() for part in module_name.split(".")):
+        raise ValueError("Invalid isolated operation module.")
+    # Explicit source identity supports installed modules and editable development
+    # checkouts without inheriting arbitrary PYTHONPATH or the parent's cwd.
+    source = Path(module.__file__).resolve()
+    source_root = source.parents[len(module_name.split(".")) - 1]
+    if source.name == "__init__.py":
+        source_root = source_root.parent
+    _check_cancelled(cancelled)
+    if admission is None:
+        with bounded_operation_admission() as reserved:
+            return run_bounded_operation(
+                operation, payload, limits=limits, cancelled=cancelled, admission=reserved,
+            )
+    _enter_admission(admission)
+    try:
+        return _run(module_name, name, source_root, payload, limits, cancelled)
+    finally:
+        admission.running = False
+
+
+def _check_cancelled(cancelled: Callable[[], bool] | None) -> None:
+    if cancelled is not None and cancelled():
+        raise ProcessBudgetError("cancelled")
+
+
+def _run(
+    module: str, name: str, source_root: Path, payload: bytes,
+    limits: ProcessLimits, cancelled: Callable[[], bool] | None,
+) -> bytes:
+    command = [
+        sys.executable, "-I", "-B", "-m", "govoplan_core.security.process_worker",
+        module, name, str(source_root), str(limits.cpu_seconds),
+        str(limits.memory_bytes), str(limits.input_bytes), str(limits.output_bytes),
+        str(limits.file_bytes),
+    ]
+    deadline = time.monotonic() + limits.wall_seconds
+    try:
+        process = subprocess.Popen(  # noqa: S603 - fixed interpreter/bootstrap, trusted operation.
+            command, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
+            bufsize=0, close_fds=True, start_new_session=True, cwd="/",
+            env={
+                "LANG": "C.UTF-8", "LC_ALL": "C.UTF-8", "TZ": "UTC",
+                "OPENBLAS_NUM_THREADS": "1", "OMP_NUM_THREADS": "1",
+                "MKL_NUM_THREADS": "1", "NUMEXPR_NUM_THREADS": "1",
+            },
+        )
+    except OSError as exc:
+        raise ProcessBudgetError("unavailable") from exc
+    output = bytearray()
+    sent = 0
+    stderr_bytes = 0
+    try:
+        with selectors.DefaultSelector() as selector:
+            for stream in (process.stdin, process.stdout, process.stderr):
+                os.set_blocking(stream.fileno(), False)
+            selector.register(process.stdout, selectors.EVENT_READ, "stdout")
+            selector.register(process.stderr, selectors.EVENT_READ, "stderr")
+            if payload:
+                selector.register(process.stdin, selectors.EVENT_WRITE, "stdin")
+            else:
+                process.stdin.close()
+            while True:
+                _check_cancelled(cancelled)
+                remaining = deadline - time.monotonic()
+                if remaining <= 0:
+                    raise ProcessBudgetError("timeout")
+                # WNOWAIT keeps the owned group leader's PID reserved until the
+                # group is killed, including descendants that close their pipes.
+                exited = os.waitid(os.P_PID, process.pid, os.WEXITED | os.WNOHANG | os.WNOWAIT)
+                if exited is not None and not selector.get_map():
+                    break
+                events = selector.select(min(remaining, 0.05))
+                for key, _event in events:
+                    if key.data == "stdin":
+                        try:
+                            sent += os.write(key.fd, memoryview(payload)[sent:sent + 65536])
+                        except BrokenPipeError:
+                            sent = len(payload)
+                        except BlockingIOError:
+                            continue
+                        if sent == len(payload):
+                            selector.unregister(key.fileobj)
+                            key.fileobj.close()
+                        continue
+                    available = (
+                        limits.output_bytes - len(output)
+                        if key.data == "stdout" else _STDERR_LIMIT - stderr_bytes
+                    )
+                    try:
+                        chunk = os.read(key.fd, min(65536, available + 1))
+                    except BlockingIOError:
+                        continue
+                    if not chunk:
+                        selector.unregister(key.fileobj)
+                    elif len(chunk) > available:
+                        raise ProcessBudgetError("output_limit")
+                    elif key.data == "stdout":
+                        output.extend(chunk)
+                    else:
+                        # Never expose raw exception/log output to a client.
+                        stderr_bytes += len(chunk)
+        exit_code = os.waitstatus_to_exitcode(
+            (exited.si_status << 8) if exited.si_code == os.CLD_EXITED else exited.si_status
+        )
+    finally:
+        # Do not communicate(): it would collect unbounded output during cleanup.
+        try:
+            os.killpg(process.pid, signal.SIGKILL)
+        except ProcessLookupError:
+            pass
+        finally:
+            for stream in (process.stdin, process.stdout, process.stderr):
+                stream.close()
+            process.wait()
+    if exit_code != 0:
+        code = {
+            71: "memory_limit", 72: "input_limit", 73: "output_limit",
+            74: "unavailable", -signal.SIGXCPU: "cpu_limit",
+            -signal.SIGXFSZ: "output_limit",
+        }.get(exit_code, "worker_failed")
+        raise ProcessBudgetError(code)
+    return bytes(output)