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 use WebFetch and web search to load public third-party documentation (e.g., https://modelcontextprotocol.io/llms-full.txt and raw.githubusercontent.com README URLs) and to read API docs as required parts of its workflow, allowing untrusted external content to directly influence the agent's decisions and tool design.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent at runtime to WebFetch external specification and SDK documents (e.g., https://modelcontextprotocol.io/llms-full.txt and https://raw.githubusercontent.com/modelcontextprotocol/python-sdk/main/README.md), which would be fetched and then used to drive prompts/instructions for MCP behavior.