managed-model-endpoints
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external model endpoints during health checks and inference smoke tests. 1. Ingestion points: Inference response data captured in Step 5 of SKILL.md. 2. Boundary markers (absent): The instructions do not define delimiters or warnings to ignore embedded instructions in model responses. 3. Capability inventory: The skill performs endpoint lifecycle management (start/stop) and inference routing defined in SKILL.md. 4. Sanitization (absent): No sanitization or validation logic is defined for the external data ingested.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to implement and execute start, stop, and health check scripts for model lifecycle management as described in SKILL.md.
Audit Metadata