grilling
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions allow user-provided answers to drive automated fact-finding missions in the local environment, presenting a potential surface for injection attacks.\n- Ingestion points: User responses provided during the structured interview process (SKILL.md).\n- Boundary markers: The instructions do not specify any delimiters or safety prompts to isolate user input from the agent's command logic.\n- Capability inventory: The skill includes the ability to dispatch sub-agents to read the filesystem and use system tools (SKILL.md).\n- Sanitization: There is no mention of sanitizing user input or validating its intent before the agent attempts to resolve facts from the environment.
Audit Metadata