learn
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages knowledge persistence through a structured workflow and emphasizes user approval for durable changes.
- [PROMPT_INJECTION]: The skill contains an inherent surface for indirect prompt injection as it updates future agent instructions based on session data. 1. Ingestion points: Session history and project file content. 2. Boundary markers: Workflow requires destinations to be proposed and approved. 3. Capability inventory: Uses Bash, Read, Write, and Edit tools for file management. 4. Sanitization: Mitigated by mandatory human-in-the-loop approval via the AskUserQuestion tool before any durable rules are modified.
Audit Metadata