Audit MCP servers for tool poisoning, hardcoded secrets, command injection, and supply-chain risks before your AI agents run untrusted code.