burner-agents-privacy

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime browsing path can ingest outsider-authored free text: the agent uses IsolatedContext to page.goto(...) arbitrary web pages and then the ReasoningEngine performs analyze_page(page) / reason_and_act(...), which requires reading the page’s visible/body text into the LLM context (public web content is an outsider source).