api-to-mcp
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill enables the creation of tools that process data from external APIs, creating a potential surface for indirect prompt injection. * Ingestion points: API response data is ingested via the _request function in the generated server.py. * Boundary markers: The generated server includes an instructions parameter providing context to the model to prevent misinterpretation of API data. * Capability inventory: The resulting servers can perform network requests (httpx) and access local token files. * Sanitization: The templates recommend slimming payloads to reduce noise and emphasize reporting null fields as unknown rather than inventing data.
- [COMMAND_EXECUTION]: The skill utilizes scripts for repository scaffolding (scaffold.py) and API reconnaissance (probe.py). These actions are transparently documented and intended for the developer workflow.
- [DYNAMIC_EXECUTION]: The scaffolding process generates Python code based on static templates. The generation logic is transparent and does not incorporate unvalidated external data into the executable code.
- [SAFE]: The skill demonstrates a strong security posture by enforcing restricted file permissions (chmod 600) for secrets, preventing sensitive data from appearing in shell history, and requiring explicit environment flags to enable write operations.
Audit Metadata