mcp-builder

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow in scripts/evaluation.py ingests the evaluation XML’s free-text <question> field (which an outsider can author/supply via eval_file) as the user prompt in agent_loop() at runtime (messages = [{"role": "user", "content": question}]), before any MCP tool selection/execution.