Verification-first accessibility engineering skills and deterministic tooling for coding agents. Provides MCP tools for accessibility scanning, evidence normalization, comparison, and verification.
Server has 5 well-named tools with complete input schemas and descriptions. Tool names follow verb_noun pattern (scan_page, scan_html, compare_reports, verify_repair, run_keyboard_journey). All tools have descriptions (50-100 chars) and detailed parameter schemas with types. However, output schemas are not documented in the provided source, only input schemas are visible. Parameter descriptions are present but vary in depth. No error handling guidance visible. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being READ_ONLY. Missing documentation of return types and structures.
Compare two audit reports to identify new, resolved, and unchanged accessibility findings
Run a bounded keyboard journey to test keyboard navigation and accessibility
Scan a static HTML file with network requests blocked for accessibility issues
Scan a rendered page for accessibility issues using axe-core
Replay a scan and verify that accessibility issues have been repaired
Output schemas not documented. Tools return audit reports and comparison results, but return type structure is not visible in source. LLMs cannot plan downstream tool calls without knowing what fields to expect.
Tool annotations missing. All 5 tools are marked READ_ONLY in metadata but lack readOnlyHint in schema. Tool annotations enable agents to reason about side effects and retry safety.
No error handling guidance. Tools accept complex parameters (viewport objects, selector arrays, storage state paths) but no documentation of validation errors, recovery steps, or what happens on invalid input.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Parameter descriptions lack constraint details. 'viewport' is an object with width/height but no min/max bounds documented. 'wcagTags' is an array but no guidance on valid tag values or limits.
compare_reports and verify_repair accept 'before'/'baseline' and 'after' report objects but no schema for report structure is provided. LLMs cannot construct valid report objects without seeing the expected format.