ai-sdk-agents

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls createMCPClient with runtime transport URLs like https://mcp-server.example.com (and https://mcp-server.example.com/sse), and then uses mcpClient.tools()/prompts()/getPrompt(), which fetch remote tools/prompts that directly control agent instructions and tool execution.