twoshot-voice-audio

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

• Third-party content exposure detected (high risk: 0.75). The workflow explicitly allows “research topic if needed (web search)” and “Search library… or import from YouTube” for voice references, which can cause the agent to ingest outsider-authored free text (e.g., scraped web pages/YouTube transcripts) into the LLM context via runtime web/video content retrieval.