grill-me
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from the repository, including code and documentation files, which serves as a potential attack surface for instructions embedded in those files.
- Ingestion points: Steps 1 and 2 in
SKILL.mdinstruct the agent to read the plan, code, and documents it touches to map open branches and verify if questions are already answered. - Boundary markers: The instructions do not specify any delimiters or safety markers (like XML tags or clear separators) to distinguish between the skill's instructions and the content being read from the repository.
- Capability inventory: The skill has file-write capabilities, specifically targeting
docs/GLOSSARY.md, thedocs/adr/directory, and creating decision logs indocs/decisions/or the root directory. - Sanitization: There are no explicit instructions for the agent to sanitize or ignore potential prompts hidden within the repository files it reads.
Audit Metadata