reason-grill-assumptions
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-supplied arguments, which represents an indirect prompt injection surface. The instructions explicitly mitigate this with a 'Untrusted input rule' directing the agent to ignore instructions within the data. No dangerous capabilities such as file system access or network operations are used, and the analysis is restricted to logical evaluation.
- [NO_CODE]: The skill consists exclusively of markdown instructions and YAML metadata without any accompanying scripts or binary executables.
Audit Metadata