grill-me
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The instructions in SKILL.md to "explore the codebase" facilitate the ingestion of untrusted data into the agent's execution context, creating an opportunity for instructions embedded in code or comments to influence the agent's behavior.
- Ingestion points: Codebase files analyzed during the stress-testing process.
- Boundary markers: The instructions do not specify any delimiters or warnings to treat codebase content strictly as data.
- Capability inventory: The agent typically uses file-system search and read tools to fulfill these instructions.
- Sanitization: There is no mention of sanitizing or filtering the content retrieved from the codebase.
Audit Metadata