autoskill

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text from the user’s screen/OCR is fetched at runtime via scripts/fetch_window.py (/search → item["content"]["text"] / window_name), then inserted into the LLM prompt through scripts/run.py → scripts/synthesize.py (cluster example titles), creating an indirect prompt-injection path.