using-model-endpoint

Pass

Audited by Gen Agent Trust Hub on Jul 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists of instructional text without any embedded scripts, tools, or binary executables, presenting no direct execution risk.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The workflow includes explicit instructions to avoid exposing secrets when saving artifacts, which is a key security measure for protecting credentials.
  • [INDIRECT_PROMPT_INJECTION]: The skill addresses the risk of untrusted data from external endpoints (Ingestion: Feynman model response) by requiring the agent to validate response shapes and use no hidden context (Sanitization/Boundaries). No specific exploitable capabilities are defined within the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 17, 2026, 04:04 AM
Security Audit — agent-trust-hub — using-model-endpoint