Repository-wide collaboration rules for humans and AI tools.
Applies to all contributions (human-authored and AI-assisted).
- Humans own all final decisions, reviews, and commits.
- AI output must be validated for correctness, security, and licensing.
- Keep changes small, traceable, and reversible.
- Do not include AI branding or AI co-author trailers in commit messages.
- Never paste production secrets, customer data, or private keys into external AI tools.
For mail pipeline changes, reviewers must verify:
- rate-limit behavior and backoff logic
- suppression list enforcement
- unsubscribe behavior
- bounce/complaint handling
- absence of credential leaks in logs
Commit messages must be human-authored. Disallowed examples include:
Co-authored-by: Claude ...Co-authored-by: ChatGPT ...Generated with Claude CodeGenerated by <AI tool/model> ...
Use scripts/check-commit-attribution.sh before pushing.