websocket-engineer

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

• Third-party content exposure detected (high risk: 0.70). The skill’s runtime workflow ingests only client-supplied WebSocket event payloads (e.g., socket.on("message", ({ roomId, text }) => ...)), which are outsider-authored free text from other users/clients, and that text is then emitted/handled by the agent’s system (indirect prompt-injection risk if the agent later forwards it into LLM context).