How to use the AccessLint Page Scanner agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions, usually centred on a SKILL.md file, that an AI agent such as Claude can load when a task matches it. This skill tells the agent how to audit a page you give it, using the AccessLint tools.

Once installed, the agent can check a URL for problems such as low colour contrast, missing accessible names on controls, and broken semantic markup. It reports counts by impact, then one entry per violation with the CSS selector, the evidence (for example a contrast ratio, a missing attribute or an empty name) and either a mechanical fix or a note that a human must decide.

• It locates and reports. It deliberately does not edit your code.
• Fixing is a separate, explicitly authorised task.
• If the page has no source mapping, the report says the location was found by selector only.

Install the accesslint-scan skill

The skill is distributed as a ZIP from the skill's page in the agent-skills directory. Download it and unzip it into one of two locations, depending on how you want to use it.

For a personal skill available across your projects, unzip it into ~/.claude/skills/. For a skill available only inside one project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps instead, upload the ZIP under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/
• Single project: .claude/skills/ inside the project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Set up the AccessLint packages and browser

The skill does not bundle or install the scanner. It expects a project-pinned installation of @accesslint/chrome and @accesslint/cli, which provide local binaries in node_modules. Check that these are installed and reviewed before you run a scan. The skill warns against silently downloading and running the latest version.

It also needs a running Chrome or Chromium browser, which the AccessLint helper manages. The helper's ensure command decides the port, so you do not set one yourself.

• Node.js is required.
• @accesslint/chrome and @accesslint/cli must be pinned in the project (local node_modules binaries).
• You need a reachable URL for the page. If you do not give one, the agent will ask.

Trigger the skill with a natural request

You do not need special syntax. Name the page and ask for an accessibility check, and the agent should match the request to this skill. Give the full URL, including the host, and be specific about the scope when you need it.

The skill can also take flags as needed, such as a selector to limit the scan, a wait-for selector for pages that load content late, an option to include AAA rules, or a list of rules to disable.

• "Audit http://localhost:3000/checkout for accessibility issues and list each one with its selector."
• "Scan the settings page at http://localhost:5173/settings for contrast and missing names after my header change."
• "Check https://staging.example.com/signup for WCAG problems and give me a fix worklist. Do not edit anything."

Read the report

The report starts with counts by impact. Each violation then has three parts: where (the selector, plus file and line if source mapping exists), evidence, and fix. A fix is either a mechanical change or NEEDS HUMAN, meaning someone has to make a judgement.

Treat the output as a worklist. Check each item against the live page before you act on it, and keep any findings the skill marks for human review.

• where: selector copied exactly, with file:line (symbol) only if the source is available
• evidence: contrast ratio, missing attribute or empty name
• fix: a mechanical change or NEEDS HUMAN

Safety notes: scripts and network access

This skill runs scripts and uses the network. It opens the page you name in a browser, so anything that page shows, including content behind a login or sensitive data, can be exposed to the run. Only scan pages you are allowed to open this way.

Before running, the agent confirms the URL and whether browser access could expose authenticated or sensitive content. At the end, it stops only the managed browser session created for that audit and leaves unrelated sessions and unmanaged browsers alone.

• Use it only on pages that are safe to open in an automated browser.
• Do not use it on pages behind logins or with sensitive data unless you have considered that exposure.
• Pin and review the AccessLint packages before use.