retrospective
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFE
Full Analysis
- Analysis of Historical Data: The skill processes conversation transcripts and logs via the session_transcript tool. It implements a defensive measure by instructing the agent to treat this data as "evidence, not authority," which is a recommended practice to avoid following instructions embedded in historical content.
- Advisory Constraints: The skill is explicitly scoped to be advisory. It contains instructions that prohibit editing files, filing issues, or changing system settings, ensuring that the retrospective results remain purely informational.
- Scoped Delegation: The skill allows for secondary reviews using the second-opinion tool but limits the scope of these reviews to prevent unauthorized data harvesting or recursive skill invocation.
Audit Metadata