grill-for-unknowns
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates as a high-level coordination framework. It instructs the agent to utilize existing tools like
researchandprototypeto gather information without executing arbitrary shell commands or downloading unverifiable external code at runtime. The URLs provided in theNOTICE.mdfile are for copyright attribution only and are not accessed during execution. - [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it processes external project data to build its "unknowns" list.
- Ingestion points: Analyzes user-provided project maps, existing source code, and documentation to identify information gaps.
- Boundary markers: Uses standardized markdown table structures to organize output, though it does not explicitly define escape sequences or delimiters for untrusted input.
- Capability inventory: Requests the invocation of other functional skills (e.g.,
research,prototype,domain-modeling). - Sanitization: There are no explicit instructions for sanitizing or validating the content extracted from the "territory" (source code/docs) before it is summarized into the agent's context.
Audit Metadata