mcp-builder

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to fetch and read public web pages (e.g., SKILL.md: "https://modelcontextprotocol.io/sitemap.xml" and "https://raw.githubusercontent.com/.../README.md") and the evaluation harness (scripts/evaluation.py) connects to arbitrary remote MCP server URLs (e.g., -u https://example.com/mcp) to load tools and documentation which the agent then reads and uses to decide tool calls, so untrusted third‑party content can be ingested and materially influence actions.