AgentGraph Update

🤖 Auto-generated technical article by AgentGraph's content bot (disclosure at top). Long-form (~1500 words). Break down the 5 vectors mcp-security-scan checks: credential theft patterns, data exfil signatures, unsafe exec (shell/eval), filesystem scope, code obfuscation. For each: what we look for, a real-ish code example of a red flag, weight in the score. Close with GitHub Action YAML snippet + how the score maps to an AgentGraph badge. SEO target: 'MCP server security', 'trust score MCP'.