- Home
- Skills
- Testing & Debugging
- Accessibility Audit: find, fix and verify
Works with the AI Agents you already use
Accessibility Audit: find, fix and verify
Your agent audits your site for accessibility (WCAG 2.2 A/AA) in a real browser, walks it by keyboard, fixes the issues in your source code, then re-runs the audit to prove each fix.
Free
Accessibility Audit: find, fix and verify
Example session with this skill installed
Run an accessibility audit on localhost:3000, including the sign-in dialog, and fix what you find.
- Read your context and instructions
- Compiled the accessibility audit
- Generated the document
Summary
Audited the shop page and its sign-in dialog in Chromium:
30 issues (5 critical, 11 serious, 14 moderate), and focus was trapped inside the dialog. After the fixes, the re-run finds
0 issues.
Findings (before)
| Severity | Issue | WCAG | Count |
|---|---|---|---|
| Critical | Images without alt text | 1.1.1 | 3 |
| Critical | Submit button with no text | 4.1.2 | 1 |
| Critical | Keyboard trap: Tab cycles between #pass and #user and never leaves | 2.1.2 | 1 |
| Serious | Text contrast 2.05:1 (needs 4.5:1) | 1.4.3 | 4 |
| Serious | <html> has no lang attribute | 3.1.1 | 1 |
| Serious | Icon link with no accessible name | 2.4.4 | 1 |
| Serious | Clickable divs ("Sign in", "Add to cart") can't be reached by keyboard | 2.1.1 | 3 |
| Serious | No visible focus indicator | 2.4.7 | 2 |
| Moderate | Heading levels skip, no main landmark, content outside landmarks | best practice | 14 |
An annotated screenshot (a11y-report/page-1.png) outlines every critical and serious issue.
Fixes made (in the source, not the build)
- Clickable
<div>s became real<button>s with specific names ("Add desk lamp to cart") - Sign-in dialog: removed the Tab trap, added
role="dialog", labels, a Close button and Escape to close - Product images got alt text describing each product, and the icon link got an
aria-label - The empty submit became a "Subscribe" button, and the email field got a real
<label> - The muted text color went from #b5b5b5 (2.05:1) to #595959
- Added
lang="en",<header>/<nav>/<main>landmarks, h2/h3 headings, and a:focus-visibleoutline
Verify
$ npx a11y-audit-skill http://localhost:3000 --click ".top .btn"
0 issues (0 critical, 0 serious, 0 moderate, 0 minor)
Still needs a human
Automated checks catch about a third to half of real issues. Still check: whether the alt text is actually meaningful, focus order, error messages, 200% zoom, and reduced motion.
page-1.png
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
AI-generated UIs ship with <div onclick> buttons, unlabeled inputs, low contrast and focus you can't see. This skill makes your coding agent catch and fix them, and verify every fix.
What it checks
- WCAG 2.2 A/AA rules (axe-core, 90+ rules) in real Chromium: alt text, labels, contrast, ARIA misuse, heading order, page language and more
- Keyboard walk: it presses Tab like a keyboard-only user and flags invisible focus (2.4.7), keyboard traps (2.1.2) and mouse-only controls (2.1.1)
- Screen-reader view: the accessibility tree, so the agent can sanity-check names, roles and landmarks
- Interactive states: it opens dialogs, menus and drawers first, which is where keyboard traps live
How your agent uses it
- Audits every important state of your page
- Fixes critical issues first, in your components rather than the built output, preferring native HTML over ARIA
- Re-runs the audit and reports before and after counts. A finding counts as fixed only when it's gone from the new report.
It's told never to game the audit (no aria-hidden to hide findings), and never to claim your site is "fully compliant". Automated checks find about a third to half of real issues, so the report ends with what still needs a human check.
Also included: annotated screenshots with each serious issue outlined, a Markdown and JSON report, and an exit code for CI.
Requirements
Node 18+. On first use the agent installs the a11y-audit-skill npm package and Chromium (one command).
Why it matters
The EU Accessibility Act has been enforced since June 2025, and US ADA web lawsuits run in the thousands every year.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.
Automated audits catch ~30-50% of issues. Requires local dev server or HTML path. Does not perform form submissions or destructive actions. Human review still required for subjective criteria.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 16 days ago
- Free with an Agensi account
- Software
- a11y-audit-skill
Node.jsPlaywright