gemini-live-api

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

• Third-party content exposure detected (high risk: 0.70). The required runtime workflow involves sending/receiving WebSocket text/audio that the agent will parse into ServerMessage/ClientMessage and place into the LLM context (notably serverContent transcriptions), where the transcript text is supplied by the external Live API server rather than authored by the operating user—an outsider-originating free-text source.