What is the web-accessibility-audit skill?
The web-accessibility-audit skill audits code against WCAG 2.0, 2.1 and 2.2. It runs four phases: automated checks, manual code inspection, prioritisation into critical, serious and moderate groups, then manual testing with keyboard and screen readers. The result is a markdown Accessibility Audit Report with next steps.
The skill's own list of the 12 most common WCAG violations opens with low colour contrast, which it puts at 86.4% of sites, citing the WebAIM Million (2021). It turns that list into an audit your agent can run on a codebase.
The four phases
- Automated: ESLint with jsx-a11y for React and JSX, the Lighthouse accessibility category, and a search for axe-core.
- Manual code inspection: grep patterns from a reference file search the code for missing alt text, keyboard navigation issues, ARIA issues, form labels, heading structure and more.
- Analyse and prioritise: findings are grouped as Critical, Serious or Moderate.
- Manual testing: keyboard use, screen readers such as VoiceOver, NVDA and JAWS, zoom, high contrast and reduced motion.
What the report looks like
The output is a markdown Accessibility Audit Report with a Summary, Critical, Serious and Moderate sections, Testing Recommendations and Next Steps. The skill warns that automated tools catch only 30 to 57% of issues, so the manual phases matter.
Who the web-accessibility-audit skill is for
Best for
- Auditing an existing front end against WCAG before a release or a client handover.
- React and JSX projects, where the ESLint jsx-a11y step applies directly.
- Teams that want findings ranked by severity, with code patterns for the fixes.
Not for
- Replacing assistive technology testing. The skill itself lists keyboard and screen reader checks as manual work.
- Legal sign-off on conformance. The report is a working audit, not a certificate.
- A short code-fix pass. See fixing-accessibility for that.
Use it with UiChemy
Audit the markup of a build first, then bring it into WordPress with UiChemy AI to WordPress. Composer can then edit the page in Elementor or Gutenberg. Click Get Started on uichemy.com. Re-test keyboard focus on the imported page. See also accessibility-compliance.
How to install Web Accessibility Audit Skill: A Four Phase WCAG Check
git clone https://github.com/warpdotdev/oz-skills.gitClone the repository
Run
git clone https://github.com/warpdotdev/oz-skills.git. The skill sits in the .agents/skills/web-accessibility-audit folder.Copy the skill folder
The README says skills live in a .agents/skills folder, either in your project or in ~/.agents/skills for a global install. Copy web-accessibility-audit into one of them.
Let the agent detect it
The README says Warp detects the skill on the agent's next interaction, so no restart step is listed.
Run an audit
Ask your agent to audit a project or a page for accessibility and point it at the source folder.
Check that it loaded
Ask the agent to audit one page. It should start with the automated phase, then inspect the code, group findings by severity and write the report.
Key features
- Three WCAG versionsThe audit covers WCAG 2.0, 2.1 and 2.2, with a POUR table and an A, AA and AAA conformance table for reference.
- Twelve common violationsThe skill lists the 12 most common WCAG violations, including low contrast at 86.4%, missing alt text at 60.6% and unlabeled form controls at 39.6%.
- Automated phasePhase one runs ESLint with jsx-a11y for React and JSX, the Lighthouse accessibility category, and a search for axe-core.
- Grep pattern inspectionPhase two uses a reference file of grep patterns to search code for missing alt text, keyboard navigation issues, colour values, ARIA issues, form labels, heading structure, language attributes, link text and media elements.
- Severity groupingPhase three sorts findings into Critical, Serious and Moderate groups so the worst problems are fixed first.
- Manual testing guidePhase four covers keyboard use, screen readers such as VoiceOver, NVDA and JAWS, zoom, high contrast and reduced motion.
- Fix patternsExamples cover alt text, contrast, captions, keyboard access, focus-visible, skip links, prefers-reduced-motion, lang, labels, error handling, ARIA, live regions and visually hidden text.
- Honest coverage noteThe skill states that automated tools catch 30 to 57% of issues and notes that pages using ARIA average 41% more errors.
- Structured reportA markdown Accessibility Audit Report lists a Summary, Critical, Serious and Moderate findings, Testing Recommendations and Next Steps.
- Reference files and scriptReferences include ARIA-patterns.md, WCAG-criteria.md, grep-patterns.md and screen-reader-guide.md, plus a run-eslint.sh script.
Frequently asked questions
What is the web-accessibility-audit skill?
It is a skill in warpdotdev/oz-skills that guides an AI coding agent through an accessibility audit against WCAG 2.0, 2.1 and 2.2. It combines automated tools, manual code inspection, severity ranking and manual testing, and it ends with a written Accessibility Audit Report.
How do I install the web-accessibility-audit skill?
Clone warpdotdev/oz-skills and copy the .agents/skills/web-accessibility-audit folder into your project's .agents/skills folder, or into ~/.agents/skills for a global install. The README says Warp detects the skill on the next interaction with the agent.
Which tools does the automated phase use?
Phase one uses ESLint with jsx-a11y for React and JSX, the Lighthouse accessibility category, and a search for axe-core in the project. A run-eslint.sh script ships with the skill. The skill notes that automated tools alone miss most issues.
How much can automated checks catch?
The skill says automated tools catch 30 to 57% of issues, and it claims about 57% automated coverage in its report. That is why it adds manual code inspection with grep patterns and manual testing with keyboards and screen readers.
What does the audit report contain?
A markdown Accessibility Audit Report with a Summary, Critical, Serious and Moderate findings, Testing Recommendations and Next Steps. Grouping by severity gives you an order for the fixes instead of one long unranked list of problems.
Which AI agents can use the web-accessibility-audit skill?
The repository README is written for Warp AI agents and Oz, and describes the .agents/skills folder. Another agent that loads SKILL.md folders from a similar path might work, but the repo does not confirm it.
Does the skill fix the problems it finds?
The skill's output is an audit and a prioritised plan. It includes code patterns for common fixes, such as skip links, focus-visible and prefers-reduced-motion, so you can ask the agent to apply them after reading the report. A related page is accessibility-review.
What license does the repository use?
GitHub lists the license as MIT. Check the license file in the repository before you redistribute the skill folder inside a product. The repository's most recent push on record was Aug 15, 2026.