videodb

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime capture workflow ingests WebSocket event payloads (including data.text from transcript / visual_index / audio_index) into a local JSONL file via scripts/ws_listener.py, and those readable text fields can later be fed into the agent’s LLM context (e.g., for summarization/alert logic), making the source outsider-authored (live user speech/screen content).