run-agent-evals
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the pnpm package manager to execute test suites and evaluation harnesses. Evidence: 'pnpm exec vitest run evals/agent-tools' and 'pnpm eval:agent-tools'.
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as its primary function is to process evaluation data and test cases, which could theoretically contain instructions designed to influence agent behavior. 1. Ingestion points: local 'evals/agent-tools' directory containing test cases. 2. Boundary markers: mentions 'native-tool boundaries' in docs/testing.md. 3. Capability inventory: execution of local evaluation scripts via pnpm. 4. Sanitization: not defined in the skill instructions; managed by the evaluation harness.
Audit Metadata