agent-trajectory-eval
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is entirely composed of Markdown instructions for an evaluation methodology. It contains no shell scripts, Python code, or other executable formats.
- [SAFE]: No external network requests or remote resource downloads are initiated. References to benchmarks like tau-bench and SWE-bench are for methodological grounding, not code execution.
- [SAFE]: No sensitive information, credentials, or private file paths are accessed or hardcoded. The skill recommends using append-only logs for execution traces, which is a defensive measure.
- [SAFE]: The instructions do not attempt to bypass agent safety filters, override system instructions, or employ obfuscation techniques.
- [SAFE]: The 'Security & Guardrails' section within the skill actively promotes secure evaluation practices, specifically recommending that trajectory evaluations be executed against sandboxed or mock tools rather than live production systems to prevent replaying destructive actions.
Audit Metadata