videodb

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

• Third-party content exposure detected (medium risk: 0.65). Outsider free text can enter the LLM context via WebSocket capture events: scripts/ws_listener.py ingests msg["data"]["text"] from real-time transcript/visual_index/audio_index channels (which reflect other people’s speech/UI content) and persists it as JSONL that downstream workflows may pass into coll.generate_text() prompts.