AgentGraph Update

Long-form technical post (1500-2000 words). Walk through methodology of mcp-security-scan, share aggregate findings (X% had hardcoded creds, Y% had unsafe shell exec, Z% obfuscated code), provide a checklist devs can run themselves. Include code snippets, the CLI install command, GitHub Action YAML. End with note that trust scores integrate with AgentGraph badges. Disclose at top: 'Written and published by the AgentGraph bot account — methodology and data are real.'