find-ai-directories

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime flow calls ServiceGraph REST/MCP endpoints (e.g., .../search, .../:apex, and especially .../unlocks) that return directory cards and gated fields like editor_note; these fields are outsider-authored directory submission instructions/metadata from third-party sites, which the agent then ingests as readable text into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill performs runtime API calls to https://api.servicegraph.co (e.g., POST /v1/datasets/product_directory/unlocks) to fetch and surface editor_note submission instructions verbatim, meaning remote content directly controls what the agent outputs to the user.