recipe-full-setup

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly accepts public HTTPS URLs as inputs and instructs (Decision Tree: "Does the input come from a public HTTPS URL? YES → Pass the URL directly to the API") to pass arbitrary third‑party-hosted images/videos to the Runway API, meaning the agent will ingest and act on untrusted external content that can influence tool behavior.