council
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests 'relevant context' from the project environment and passes it directly to multiple sub-agents via the
delegate_tasktool. There are no instructions for sanitizing this input or using boundary markers (like XML tags or 'ignore instructions' headers) to separate data from instructions. This creates a surface where malicious content embedded in the codebase being reviewed could influence the sub-agents' behavior. - Ingestion points: Project context and user request passed to delegates.
- Boundary markers: Absent in the instructions provided to delegates.
- Capability inventory: Use of
delegate_taskandsubagent toolsto trigger reasoning and review tasks. - Sanitization: No validation or filtering of the external context is performed before delegation.
Audit Metadata