fastmcp

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). Outsider free text can enter the LLM context via context.request_elicitation(...) (user-provided responses) and via context.request_sampling(...) (messages passed to the LLM), both of which are runtime text originating from parties the operating user didn’t author.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The OpenAPI integration code loads an external spec at runtime via httpx.get(OPENAPI_SPEC_URL) (defaulting to https://api.example.com/openapi.json) and passes that spec into FastMCP.from_openapi to auto-generate the server's tools/resources (i.e., content fetched at that URL directly controls the agent's exposed prompts/tools).