Build this before asking an agent to implement.
What outcome are we trying to create?
- Repo:
- Stack:
- Relevant subsystems:
- Test/build commands:
What recurring problem, friction, or risk triggered the work?
Hard constraints:
Soft constraints:
Assumed constraints to question:
- failing test:
- repro:
- logs:
- support/report link:
- review comment:
Stop if:
Run dissent if:
Salvage if:
Where did this context come from?