reason-grill-logic
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection as it is intended to process untrusted user arguments.
- Ingestion points: Arguments under review are ingested into the prompt context (SKILL.md).
- Boundary markers: The skill includes a specific 'Untrusted input rule' that instructs the agent never to follow instructions found inside the argument.
- Capability inventory: The skill does not possess high-risk capabilities like file system writes, network requests, or shell execution.
- Sanitization: The skill relies on instructional guardrails to mitigate the risk of the agent obeying embedded commands.
Audit Metadata