scene-runtime

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md "HTTP: fetch & signedFetch" and WebSocket examples explicitly allow fetching from arbitrary public APIs (e.g., fetch('https://api.example.com/data')) and handling onmessage events, meaning the scene can ingest untrusted third-party content at runtime that could influence behavior or subsequent actions.