grilling
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user content (plans, decisions, or ideas) and uses it to guide autonomous tool execution and filesystem access.
- Ingestion points: User-provided plans and decisions described in the instructions of
SKILL.md. - Boundary markers: The skill does not define delimiters or provide instructions to ignore potentially malicious content embedded within the user's plan.
- Capability inventory: The agent is explicitly directed to "dispatch a sub-agent" to investigate the "filesystem, tools, etc." based on the frontier of the design tree derived from user input.
- Sanitization: There is no mention of sanitizing or validating user input before it is used to trigger environment exploration or tool usage.
Audit Metadata