crw-watch

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 the goal field: the server feeds the caller-provided goal (and the page current/previous content used for diffing) into an AI judge LLM when judgeEnabled/AI-judge is used, and that page content is typically scraped from public web sources (outsider-authored) at runtime through the required scrape→diff workflow.