council
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, command execution, or unauthorized network operations were detected. The skill is designed to guide the model through a multi-perspective decision council workflow.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined questions and codebase context, which are then passed to subagents. The instructions mitigate potential injection by requiring fresh subagent contexts with minimal history, serving as a boundary against conversational anchoring or instruction leakage from the main thread.
- [PERSISTENCE]: The skill explicitly forbids writing to non-standard or hidden paths, such as shell profiles or local notes directories. It recommends using established project tools like GitHub or Linear for storing outcomes, ensuring transparency and standard operational practices.
Audit Metadata