AccessLint Skills - Setup Guide¶

Source: AccessLint/skills (98⭐, created Oct 2025, pushed Aug 25, 2026) Skill: AccessLint/skills (5 installable skills; skills.sh indexes 13 listings - 5 core + 8 legacy aliases) Combined Installs: ~5.1K across indexed listings (Sep 15, 2026 snapshot) Category: Web Accessibility / Testing Quality Tier: 🟡 Unverified (no skills.sh Trust Hub / Socket / Snyk verdicts published - verified Sep 15, 2026)

AccessLint is an established web-accessibility vendor (npm @accesslint/cli, @accesslint/core). Its skills repo ships five Agent Skills that cover the WCAG 2.2 audit workflow end to end: automated scanning, hands-on keyboard and screen-reader checks, WCAG-EM conformance auditing, remediation, and regression diffing. Each finding is graded on two axes - severity (user impact) and evidence basis (● verified / ◐ confirm-with-a-human / ○ human-required) - making this the most audit-rigorous a11y skill set currently on skills.sh. Install is agent-agnostic via the skills CLI; a bundled MCP server (@accesslint/mcp) is optional.

Installation¶

# All 5 skills (verified live Sep 15, 2026: "Found 5 skills")
npx skills add AccessLint/skills

# Optional: AccessLint MCP server (rule metadata for audit/inspect/fix)
# mcpServers entry:
#   "accesslint": { "command": "npx", "args": ["-y", "@accesslint/mcp@latest"] }

A Claude Code plugin variant exists (claude plugin marketplace add accesslint/skills), but the skills CLI path is the agent-agnostic one.

Prerequisites¶

Requirement	Details
Node.js 18+	`npx skills` requirement; `@accesslint/cli` is shelled out via npx on demand
Chrome DevTools MCP	`accessibility-inspect` drives a live page via `mcp__chrome-devtools__*` tools
`accesslint.config.json`	Optional: `npx @accesslint/cli init` defines named targets so skills can run without a URL
AccessLint MCP (optional)	`list_rules` / `explain_rule` metadata for audit/inspect/fix

Roster - 5 Core Skills¶

Skill	Installs	Scope	Does
accessibility-audit	576	whole site	WCAG-EM: defines scope, samples pages/flows, runs scan+inspect per page, reports per-criterion conformance (pass/fail/undetermined)
accessibility-scan	518	one page	Automated rule engine against the live DOM; returns a worklist (selector, file:line, evidence, fix directive). Locates; doesn't edit
accessibility-fix	510	target/worklist	Baseline → edit → verify; applies mechanical fixes, leaves TODOs for visual/contextual judgment
accessibility-inspect	488	one page	Manual tier: keyboard/focus order, a11y-tree names/roles/states, reflow/zoom at 200%, reduced motion, form errors
accessibility-diff	485	a change	Diffs a page against a baseline (uncommitted changes or a branch); reports only what the change introduced/fixed

skills.sh also indexes 8 legacy/alias listings (audit 654, scan 399, diff 365, contrast-checker 287, refactor 258, use-of-color 246, link-purpose 242, audit-and-fix 30) - the repo itself ships the 5 skills above (live-verified with --list).

CorpusIQ Use Cases¶

Use Case	How
Docs-site compliance	Run `accessibility-audit` against corpusiq.io/docs pages ahead of deploy; apply mechanical fixes with `accessibility-fix`
Section 508 / enterprise client work	WCAG-EM conformance reports for government or enterprise engagements
Product UI checks	`accessibility-inspect` keyboard/focus-order passes on new UI templates before release
Regression guard	`accessibility-diff` on docs PRs - reports only what a change introduced or fixed
Help-first content	Real a11y findings become operator-help content (real problems, real fixes)

Limitations / Verification¶

accessibility-scan / -diff work immediately (shell out to @accesslint/cli); -fix / -audit / -inspect benefit from the optional MCP server
The methodology draws an explicit boundary against standing in for real assistive-technology users (○ human-required findings) - don't over-claim compliance from agent runs alone
Verify: npx skills add AccessLint/skills --list shows "Found 5 skills"; then npx @accesslint/cli --version

Security¶

No skills.sh security audits published (verified Sep 15, 2026 - the publisher page renders no audit verdicts). Treat as unverified until reviewed:

Audit	Verdict
Gen Agent Trust Hub	Not published
Socket	Not published
Snyk	Not published

← Skills Catalog | Marketplace → Powered by CorpusIQ