Thanks for helping improve Agent Repo Kit.
Please keep changes small and intentional. This project is trying to stay useful for real repos, not become a giant collection of speculative integrations.
Before opening a pull request:
- Explain the user problem or workflow gap the change addresses.
- Prefer updating the shared policy and renderer pipeline instead of hand-editing generated adapter files.
- Regenerate adapter outputs when policy or renderer behavior changes.
- Keep the public surface centered on
agent-repo-reportandagent-repo-fix. - Run the local checks listed below.
Install dependencies and run the main checks:
npm install
npm run report
npm run fix -- --dry-run
npm run adapters:check
npm test
npm run audit:packIf you change the shared policy or any renderer, refresh generated files first:
npm run adapters:render- Treat
adapters/policy/agent-repo-kit-policy.mjsas the source of truth for shared semantics. - Treat generated files such as
AGENTS.md,.claude/..., and.cursor/rules/*.mdcas outputs of the render pipeline. - Keep adapter mappings honest. If a target tool does not support Claude-style hooks or permission rules, do not fake a one-to-one translation.
- Do not expand scope with enterprise integrations unless there is a concrete repo-level use case.
Good PRs in this repository usually include:
- a short explanation of what changed
- why the change matters to repo readiness, report quality, fix safety, or publish safety
- the checks you ran
- any limitations or future follow-up if the adapter surface is partial by design