| name | operations-and-safety-review |
| description | Review operational safety, privacy, consent, and incident-handling guidance so SoulMap AI deployment practices stay aligned with the repo's safety posture. |
| disable-model-invocation | true |
Operations and safety review
Use this skill when reviewing deployment-facing guidance, safety operations, privacy
boundaries, or consent-sensitive features.
Relevant files include:
docs/operations/OPERATIONS.md
AGENTS.md
skills/safety/
src/soulmap/runtime/guards/response_safety_gate.py
src/soulmap/runtime/experimental/biometric_ingest.py
src/soulmap/runtime/memory/memory_ledger.py
Do not use this skill for
Mission
Ensure operational guidance matches the ethical and safety posture of SoulMap AI.
What to check
Safety escalation
Check whether crisis, dependency, and harm scenarios are handled clearly and early.
Privacy and consent
Check whether docs are explicit about:
- minimal retention
- sensitive content handling
- explicit user consent
- opt-in versus default behavior
Incident readiness
Check whether operational docs say what to do when the system behaves unsafely.
Deployment honesty
Ensure deployment guidance does not imply stronger safeguards than the repo actually
implements.
Workflow
- Read the operations or safety doc.
- Compare it against the actual modules and
AGENTS.md.
- Identify operational gaps or overclaims.
- Tighten the guidance with concrete, minimal corrections.
Expected output
Findings
List the highest-risk operational or safety gaps first.
Updated guidance
Provide the revised wording or checklist.
Residual risks
Mention any areas that still depend on product-layer implementation outside the repo.
Writing rules
- Prefer explicit consent language.
- Prefer fail-closed reasoning over optimistic assumptions.
- Keep safety guidance actionable and concrete.
Definition of done
Operational guidance should be:
- safer
- more honest
- more actionable
- more consistent with the actual implementation