research

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs agents to fetch and read external web content (e.g., the "網頁抓取策略" section which prefers WebFetch and fallback to Chrome mcp__claude-in-chrome__navigate / mcp__claude-in-chrome__get_page_text, the Map Phase and Map/Reduce prompts, and custom-perspective "search" methods), so the agent will ingest untrusted public web/user-generated content at runtime and act on it, enabling indirect prompt injection risk.