tts-generation

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

• Third-party content exposure detected (high risk: 0.75). The runtime reads {workspace}/data/script.md and sends the parsed outsider-authored script text (speaker turns) as free-form prompt input to the Gemini Interactions API (client.interactions.create(... input=prompt ...)), so any non-user-authored script content becomes LLM context (indirect prompt injection risk).