From axe-accessibility
This skill should be used when the user asks to "audit accessibility", "fix all a11y issues on this page", "run the accessibility loop", "remediate accessibility until clean", "scan and fix localhost", or runs /axe-accessibility:mcp-audit. It drives the axe MCP analyze -> remediate -> apply -> verify loop on a URL, applying fixes each round until violations reach zero or a round cap is hit.
How this skill is triggered — by the user, by Claude, or both
Slash command
/axe-accessibility:mcp-audit <url> [max-rounds] (default url: detected localhost, default rounds: 5)<url> [max-rounds] (default url: detected localhost, default rounds: 5)This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Take a page from "has violations" to "zero violations" by iterating the axe MCP workflow and applying real code fixes each round. This skill assumes the axe MCP Server is already connected (`/axe-accessibility:mcp-setup`) and the `analyze` and `remediate` tools are available.
Take a page from "has violations" to "zero violations" by iterating the axe MCP workflow and applying real code fixes each round. This skill assumes the axe MCP Server is already connected (/axe-accessibility:mcp-setup) and the analyze and remediate tools are available.
$1): the full URL to audit (scheme + host + port + path). If omitted, detect a running dev server: check package.json scripts and common ports (3000, 5173, 8080, 4321) and confirm the URL with the user before scanning. Never scan a partial path.$2): iteration cap, default 5. The loop stops at zero violations or when this cap is reached, whichever comes first.If the target is behind auth or an interaction, gather the steps and pass them to analyze via its before array (click / fill / waitFor); put secrets only in fill value, never in selectors.
Repeat until zero violations or the round cap:
analyze with the full URL. Record the returned pageUrl and the list of issues. If zero issues on round 1, report "already clean" and stop.remediate once with the mapped fields:
pageUrl = analyze response pageUrlrule = issue ruleelementHtml = issue sourceremediation = issue description + " " + helpText
Deduplicate first — if the same rule + source appears many times, remediate it once and apply the fix to the shared source (each remediate call costs credits).mcp-usage skill's references/rule-tips.md.analyze on the same URL. If zero, stop and report success. Otherwise continue to the next round with the remaining issues.Track issue counts per round so progress is visible (e.g. round 1: 7 -> round 2: 2 -> round 3: 0).
remediate. Do not refactor unrelated code or change behavior beyond what the violation requires.analyze is cheap (no remediation credits); re-running remediate for an already-fixed element wastes credits — only remediate violations present in the latest analyze pass.For field-mapping details and rule-specific remediation nuances, the mcp-usage skill's references apply directly.
npx claudepluginhub dequelabs/axe-accessibility --plugin axe-accessibilityGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.