Practical Notes
- Repo stats claimed in README: 132 rules, 32 domains, 190 docs, and 6,500+ lines of code (check the README for the latest counts)
- Setup time ~10 minutes (clone + choose one rule to start)
- GitHub stars + forks (verified): see Source & Thanks
Treat rules as living policy, not a one-time paste:
- Start with one rule file that matches your stack (Next.js, Python, Rust, etc.).
- Add “safe defaults” first: no secrets, no destructive ops without confirmation, explicit file boundaries.
- After each PR, update rules based on what the AI got wrong (hallucinated APIs, ignored architecture, unsafe edits).
This library is most effective when you combine it with a repo-specific rule generator (e.g., generate a baseline, then replace/extend with curated rules).
FAQ
Q: Should I copy all rules at once? A: No. Start with one domain rule and expand only when you see repeated failure patterns.
Q: How do I know a rule works? A: Run a small PR and check whether the AI consistently follows constraints (paths, architecture, security).
Q: Can I mix with other rule systems? A: Yes. Use curated rules as overlays on top of your repo baseline guidance.