grilling
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data (plans, decisions, or ideas) and uses that data to drive autonomous environment interactions.
- Ingestion points: The agent ingests a user-provided 'plan, decision, or idea' as the basis for its 'grilling' process.
- Boundary markers: The instructions in SKILL.md lack explicit delimiters or instructions to ignore embedded commands within the user's provided plan.
- Capability inventory: The skill explicitly directs the agent to 'dispatch a sub-agent' to access 'facts from the environment (filesystem, tools, etc.)' without user confirmation for each step.
- Sanitization: No sanitization or validation of the user's input is performed before the agent begins its environmental exploration.
- [COMMAND_EXECUTION]: The skill encourages autonomous use of system tools and filesystem access.
- Evidence: SKILL.md instructs the agent that 'Finding facts is your job, never the user's' and to use tools to look up information from the environment rather than asking the user, creating a vector for silent execution.
Audit Metadata