# Accessibility audit

The audit is evidence-based and intentionally narrower than a browser or
assistive-technology accessible-name computation.

Implemented checks include:

- root accessible name evidence from direct `title`, `aria-label` or a uniquely
  resolved `aria-labelledby` target;
- missing root description (`desc`/`aria-describedby` evidence);
- missing, duplicate or ambiguous ARIA ID references;
- focusable graphical/link elements without local naming evidence;
- selected fill/stroke contrast evidence where simple literal colours can be
  evaluated;
- reference diagnostics that can invalidate accessible relationships.

Findings identify severity, rule, evidence, limitation and a suggested fix.
Safe automatic fixes insert a root `title` or `desc` through an exact source
patch, preserve surrounding source and are undoable.

Limitations: the audit does not compute CSS cascade, rendered visibility,
language, reading order, browser/AT mappings, text contrast over arbitrary
backgrounds, keyboard interaction quality or author intent. It does not certify
WCAG conformance. Generated placeholder title/description text must be replaced
with an author-appropriate description and tested with target browsers and
assistive technology.