pre-commit #

Check changed agent-instruction files as part of your normal commit workflow. Start with advisory findings, or add a severity threshold to block commits that need attention.

Setup #

With pre-commit available, add this entry to .pre-commit-config.yaml, preserving your existing hooks:

repos:
  - repo: https://github.com/hermes-labs-ai/lintlang
    rev: v0.8.0
    hooks:
      - id: lintlang

Install the repository's hooks:

pre-commit install

The default hook selects changed AGENTS.md, CLAUDE.md, GEMINI.md, and SKILL.md files at any depth; agent.yaml, agent.yml, and agent.json; .github/copilot-instructions.md; and *.instructions.md under .github/instructions/. No fixed input path is required.

Verify #

pre-commit run lintlang
pre-commit run lintlang --all-files

The first command checks matching staged files; the second checks all matching files tracked by Git. An unrelated commit can skip this hook. Findings are advisory by default, including a FAIL verdict; missing, unreadable, or unparseable inputs still fail.

Choose a blocking threshold #

Add this under the hook to block HIGH and CRITICAL findings:

      - id: lintlang
        args: [--fail-on, fail]

Use review instead of fail to include MEDIUM findings. pre-commit appends changed filenames to args, so adding a path there does not replace the default selection.

For one fixed path on every commit, use all three settings:

      - id: lintlang
        args: [AGENTS.md, --fail-on, fail]
        pass_filenames: false
        always_run: true

Next steps #

Review an existing backlog with baselines · Understand verdicts · All integrations

To remove the integration, remove only its entry from your pre-commit configuration. The released setup guide contains the full selection and argument contract.