Playbook
Write AGENTS.md
The constitution of the repo. Commands, boundaries, done. Point at skills for depth.
When: A coding agent will touch this repository.
- 01
Commands that actually run
Install, test, typecheck, build. If the agent cannot run them, they are decoration.
Verification as Architecture - 02
Boundaries
Main, secrets, prod, what never to do. Short. True. The repo itself must follow them.
AGENTS.md as Constitution - 03
Point at skills
Do not paste a UI treatise. Link the skill that should load.
Writing Skills That Trigger
Done when
- Root AGENTS.md with install/test/build, secrets rule, and a done condition the harness can run.
Refuse
- No verifier in the loop. Verify tool from AGENTS.md. Done = reviewable diff + tests + human gate.
- Unowned prompt. Store prompts in git. Read what the framework sends. Eval on change.