Risk Classification
Classify a task's risk before an agent starts, so the right review level is chosen up front.
Task summary
...
Reversibility
- [ ] Fully reversible
- [ ] Reversible with migration / rollback
- [ ] Hard to reverse
- [ ] Not reversible
Production impact
- [ ] None
- [ ] Indirect
- [ ] Direct staging impact
- [ ] Direct production impact
Sensitive areas
Does the task touch:
- [ ] Auth
- [ ] Authorization
- [ ] Crypto
- [ ] Secrets
- [ ] CI / CD
- [ ] Infrastructure
- [ ] Database schema
- [ ] Production data
- [ ] External write APIs
- [ ] Public API contracts
Proposed tier
- Tier 1 / Tier 2 / Tier 3
Required review
- None / Normal / Rigorous / Explicit human approval
Reasoning
...
Decision owner
...