find-mcp-directories

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

• Third-party content exposure detected (high risk: 0.72). The skill’s runtime calls the ServiceGraph product_directory dataset (e.g., GET .../search and optional POST .../unlocks) to ingest directory metadata and gated fields like editor_note/traffic into the agent context; these directory records are outsider-authored catalog content (not authored by the operating user), creating indirect prompt-injection risk via free-text fields returned by the API.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime calls to https://api.servicegraph.co (e.g., /v1/datasets/product_directory/... and /unlocks) and surfaces returned gated fields like editor_note verbatim, so externally fetched content can directly control the agent's output/instructions.