search

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

• Third-party content exposure detected (high risk: 0.85). The required workflow uses Tavily web search (public web pages) via the Tavily MCP server, and when results are returned with extracted page text (e.g., content and optionally include_raw_content), that outsider-authored free text is ingested into the agent’s LLM context through the MCP tool output.