testing-tool-evaluator
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill's Python code uses the 'requests' library to perform GET requests to external 'api_endpoint' URLs for performance benchmarking. This constitutes network activity to non-whitelisted domains.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection because it is designed to fetch and process content from external tools and configurations.
- Ingestion points: Data is ingested from the 'tool_config' structure and results from external API calls.
- Boundary markers: The instructions do not define boundaries or provide guidance to the agent on how to separate untrusted tool data from core instructions.
- Capability inventory: The skill possesses network communication capabilities via 'requests' and data processing features through 'pandas'.
- Sanitization: The evaluation framework contains no logic for sanitizing or validating external input before it is evaluated by the agent.
Audit Metadata