Chat
Pass
Audited by Gen Agent Trust Hub on Apr 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a localized preference-tracking system with well-defined boundaries and transparency rules. It does not perform network operations or access sensitive system areas.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill creates a local directory (~/chat/) to store preference files. This is consistent with its stated purpose and does not involve data exfiltration or access to unauthorized file paths like SSH keys or credentials.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user feedback to adapt its behavior, which creates a potential surface for indirect influence. 1. Ingestion points: Preferences enter via user feedback and are read from ~/chat/memory.md (SKILL.md). 2. Boundary markers: Not explicitly defined for the storage files. 3. Capability inventory: File system read/write access limited to the ~/chat/ directory. 4. Sanitization: The skill uses a 'Three-Strike Confirmation' process and strict formatting rules defined in criteria.md to validate and sanitize patterns before they are committed to memory.
Audit Metadata