mcp-builder

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

• Third-party content exposure detected (high risk: 0.90). The guide explicitly instructs the agent to fetch and read public web content (e.g., "https://modelcontextprotocol.io/sitemap.xml", raw GitHub URLs like "https://raw.githubusercontent.com/modelcontextprotocol/.../README.md", and "use web search and WebFetch as needed"), so the agent will ingest untrusted public third‑party content as part of its workflow, enabling potential indirect prompt injection.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The evaluation harness accepts a runtime MCP server URL (e.g., https://example.com/mcp) for SSE/HTTP transports and will call tools on that remote server during evaluation, which executes remote code and can directly control the agent's actions and prompts.