five-flavors-balance
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions and metadata were analyzed for all 10 threat categories. No evidence of credential exposure, data exfiltration, or unauthorized code execution was found.
- [PROMPT_INJECTION]: Analyzed the attack surface for indirect prompt injection. Ingestion points: User dietary descriptions and health symptoms (SKILL.md). Boundary markers: Explicit 'Boundary' section defines exclusions for medical emergencies and professional plans. Capability inventory: Confirmed the absence of subprocess calls, file system modifications, or network operations across all files. Sanitization: The skill's logic is purely informational and lacks dangerous execution sinks, mitigating risks from untrusted user input.
Audit Metadata