agent-eval

Pass

Audited by Gen Agent Trust Hub on Apr 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the 'judge' configuration in YAML tasks to execute arbitrary shell commands (such as pytest or npm run build) for verifying agent output.
  • [EXTERNAL_DOWNLOADS]: Provides a reference to a GitHub repository (github.com/joaquinhuigomez/agent-eval) for the underlying agent-eval CLI tool.
  • [PROMPT_INJECTION]: The skill processes external YAML task definitions containing prompts and commands; while intended for evaluation, this represents a surface where untrusted data could influence agent actions.
  • Ingestion points: YAML task definitions in the tasks/ directory.
  • Boundary markers: None explicitly defined in the task schema.
  • Capability inventory: Uses Bash, Write, and Edit tools to execute tests and apply agent-generated code.
  • Sanitization: No specific sanitization or validation of the task content is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 14, 2026, 12:19 AM
Security Audit — agent-trust-hub — agent-eval