01 / Quick start

Run once with npx.

Use this when you want to audit the current repo without changing package.json. AgentLint scans the harness files it can find and reports issues against the 33-check catalog.

npx agentlint

02 / Project install

Add AgentLint to dev dependencies.

A local install makes the linter version part of the project. That is the better default for teams because CI, local scripts, and future contributors run the same package version.

npm install --save-dev agentlint

Add a package.json script.

Keep the command discoverable. A named script also gives agents a stable command to run when they need to verify harness changes.

{
  "scripts": {
    "agentlint": "agentlint",
    "lint:harness": "agentlint"
  }
}

03 / GitHub Actions

Check pull requests.

Put AgentLint in CI when changes to CLAUDE.md, AGENTS.md, Cursor rules, hooks, or workflows should be reviewed before merge. This workflow runs on every pull request.

.github/workflows/agentlint.yml

name: AgentLint

on:
  pull_request:

jobs:
  agentlint:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - run: npx agentlint

04 / Pre-commit hook

Catch staged harness changes early.

A pre-commit hook is useful when harness files change often. Keep CI as the source of truth, but use the local hook to catch obvious stale paths, contradictions, and missing verification before review.

npx husky init
echo "npx agentlint --staged" > .husky/pre-commit
chmod +x .husky/pre-commit

Review the AgentLint source and package notes.

Use the repository for release notes, issues, and the latest CLI details before pinning AgentLint in a larger team workflow.

Open agentlint GitHub