| name | safety-md-incident-update |
| description | User-invoked skill for updating SAFETY.md hazard logs and safety evidence after an incident, near miss, or safety signal. |
| disable-model-invocation | true |
SAFETY.md Incident Update
Use only when explicitly asked to update safety documentation after an incident, near miss, complaint, safety signal, or post-market learning.
Boundary
This skill does not replace organisational incident reporting, statutory reporting, clinical governance, MHRA vigilance, IG breach reporting, or duty-of-candour processes. It updates repository-native safety evidence.
Steps
- Capture incident facts without inventing missing details. Completion: timeline, affected users/patients/systems, harm/near-miss status, and unknowns are explicit.
- Link to existing hazards or propose new hazards. Completion: every incident mechanism maps to an H-ID or proposed new hazard.
- Update controls/evidence needs. Completion: failed controls, missing controls, and verification evidence are clear.
- Identify safety case and release/configuration implications. Completion: affected sections/releases are listed.
- Record questions/actions for CSO and responsible owners in
queries.md or an action list. Completion: required human decisions are explicit.
Never do
- Do not minimise harm or residual risk.
- Do not mark hazards closed because an incident has been documented.
- Do not determine reportability unless qualified evidence is present; flag for responsible review.
Completion criteria
- Incident is traceable to hazards/controls.
- Hazard log update or proposed update exists.
- Evidence and owner gaps are clear.
- CSO/regulatory/IG review prompts are explicit where relevant.