grill
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill functions entirely as a set of logical instructions for agent dialogue. It does not include scripts, binaries, or configuration that would enable command execution or network access.
- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to gather context by reading local repository files and documents. While this introduces a data ingestion surface, the risk is effectively mitigated by the skill's restrictive scope.
- Ingestion points: SKILL.md instructs the agent to 'Read relevant repository files, documents, tests, or available sources' to bind uncertainty.
- Boundary markers: No explicit markers are defined to delineate external content from the skill's internal instructions.
- Capability inventory: The skill explicitly prohibits implementation ('Grill does not implement the result') and contains no subprocess, network, or file-write capabilities.
- Sanitization: There is no evidence of specific sanitization or filtering logic applied to the ingested content.
Audit Metadata