evaluate-skill

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute specific Node.js scripts located in the project structure (".agents/skills/platform/evaluate-scenarios.mjs" and "evaluate-behavioral-fixtures.mjs") to perform regression testing and behavioral validation.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from scenario and behavioral fixtures, which are external ingestion points that could contain instructions designed to influence the agent's evaluation output.
  • Ingestion points: Files loaded from the "scenarios/" and "behavioral-fixtures/" directories (SKILL.md).
  • Boundary markers: The Contract section defines that the skill should only produce reasoning or documentation artifacts and not modify external systems unless explicitly instructed.
  • Capability inventory: The skill has the capability to execute local scripts and generate Markdown artifacts.
  • Sanitization: There are no mentioned mechanisms for sanitizing or escaping the content within the scenario fixtures before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:00 AM
Security Audit — agent-trust-hub — evaluate-skill