When a rendered page is available, the skill instructs the agent to first run a Lighthouse accessibility audit (calling Chrome DevTools MCP directly if available), locate the corresponding component or template based on failed nodes, inspect the rendered accessibility tree, test affected workflows using keyboard navigation, and rerun the same audit after making fixes; when these tools are unavailable, it falls back to the Lighthouse CLI or axe. The main text is based on WCAG 2.2, listing rules and positive/negative examples for alternative text, contrast, focus, form labels, etc., according to the four principles of perceivability, operability, understandability, and robustness, while reminding that a perfect score does not equal WCAG compliance.
Why it matters Requiring agents to first run real audits, inspect the accessibility tree, and test workflows via keyboard navigation before modifying source code based on failed nodes helps avoid blind changes across the entire repository.
Security Audit: Not audited
- Not yet scanned
Rationale
- skills.sh 24 hours installed 443 (only indicates attention level); SKILL.md is marked MIT
- Explicitly states that automated tools can only detect a portion of accessibility issues and must be combined with manual checks
Usage
- Run npx skills add https://github.com/addyosmani/web-quality-skills --skill accessibility, then tell the agent "Perform an accessibility audit on the homepage, and rerun Lighthouse after fixing issues"
Risks
- Not audited: Desktop version not scanned, website does not display installation commands
- The complete workflow requires a browser or Chrome DevTools MCP; skills.sh currently lacks cumulative installation counts