Commit format that AI agents and humans can read, understand, and act on.
type(Scope): what (why) → next
feat(AuthService): add JWT validation (token expiry protection)
wip(AuthController): add logout (security) → token blacklist, rate limiting
fix(SessionManager): validate user ID (users crashed on empty session)| Capability | Reads | Does |
|---|---|---|
| Resume | → next |
Continue after crash |
| Review | (why) |
Explain past decisions |
| Handoff | Full history | Summarize for new agent/developer |
| Code Review | (why) + diff |
Evaluate if solution fits problem |
Add to your CLAUDE.md or AGENTS.md:
Commit format: type(Scope): what (why) → next
- type: feat/fix/wip/refactor/test/docs/chore
- Scope: file name or component
- (why): motivation
- → next: continuation (wip only)
Or install the full skill from the skills/agentic-commit directory.
MIT