grilling
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (plans, decisions, or ideas) and instructs the agent to automatically gather "facts" from the environment without user intervention. This creates a risk where malicious input could trick the agent into accessing sensitive files or executing tools via the sub-agent mechanism.
- Ingestion points: User-provided descriptions of a "plan, decision, or idea" defined in SKILL.md.
- Boundary markers: Absent; there are no instructions to the agent to treat user data as untrusted or to ignore instructions embedded within the user's plan.
- Capability inventory: SKILL.md explicitly instructs the agent to use "filesystem, tools, etc." by dispatching sub-agents to find facts.
- Sanitization: Absent; no validation or escaping of the user's input is mentioned before it is used to determine which environment facts to look up.
Audit Metadata