autoskill

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

• Third-party content exposure detected (medium risk: 0.30). The autoskill workflow fetches screen OCR and window title text from the local screenpipe daemon via loopback HTTP (http://localhost:3030/search), which captures the user's active screen activity (potentially including outsider-authored content from browser windows, chat feeds, or documents) and passes it into the LLM synthesis pipeline.