paseo-committee
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied conversation data to formulate prompts for secondary agents instantiated via the platform's orchestration tools.\n
- Ingestion points: The instructions in SKILL.md specify that the user's request and current conversation are used as input for the committee members.\n
- Boundary markers: The skill implements a 'no-edits suffix' (restricting file modification and code writing) which acts as a behavioral boundary, though it does not use structured data delimiters for the untrusted input.\n
- Capability inventory: The skill leverages the create_agent and list_profiles tools to spawn and manage high-reasoning agent instances.\n
- Sanitization: There is no evidence of sanitization or escaping of the conversation content before it is interpolated into the prompts for the committee agents.
Audit Metadata