ag2-eval-comparison
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'ag2' Python package. This is a legitimate dependency provided by the skill vendor for performing agent evaluations.
- [COMMAND_EXECUTION]: The documentation includes a 'pip install' command to set up the necessary environment for running the evaluation scripts.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes task data from a 'suite' without explicit boundary markers or sanitization. Malicious content within the evaluation tasks could theoretically influence the agents or the judges being compared.
- Ingestion points: The 'suite' object passed to 'run_variants', 'run_pairwise', and 'evaluate_pairwise' functions.
- Boundary markers: No specific delimiters or instructions to ignore embedded commands are used when interpolating task data into the evaluation flow.
- Capability inventory: While the evaluation skill itself is safe, the 'Agent' instances it executes may have broad capabilities depending on their specific implementation and configuration.
- Sanitization: The instructions do not include steps to sanitize or validate the content of the evaluation suite before processing.
Audit Metadata