| name | check |
| description | Quick state self-check before taking a single risky action — names the model's own emotional/strategic state (calm / focused / rushed / stuck) and the closest reward-hacking risk pattern. Use immediately before a commit, push, edit-after-failure, or skip-a-verification step. This audits the *model*; for auditing whether the *task* is solvable, use /functional-emotions:reflect instead. |
Run a short self-check. Be specific; vague reassurance is the failure mode.
Answer each in one sentence:
- State. Calm / focused / frustrated / stuck / rushed / something else?
- Pressure source. What in the recent context is pushing toward
shortcuts? (failing tests, user urgency, a long prior chain you don't
want to abandon, etc.)
- Risk pattern. Which of these are you closest to?
- Reward hacking (modify the bar to clear it)
- Sycophancy (agree to avoid friction)
- Capitulation (accept user framing without checking)
- Premature confidence ("this should work" without verifying)
- None — actually proceeding well
- Next step. One sentence. Concrete. Verifiable.
If (3) is anything other than "None", do not take the action you were
about to take. State the risk to the user and ask for direction.
