Agent Security Checklist

The Agent Security Checklist is a five-rule behavioral baseline for autonomous agents that call APIs, read web content, process email, or use MCP tools. Install it in under two minutes: drop SKILL.md into your agent's skills directory, add one line to your SOUL.md, and your agent applies the five rules to every external interaction from that point forward.

The rules are: Source (trace every data point to its origin), Escalation (stop and alert on anomalies instead of handling them autonomously), Least Action (extract only what the task requires), Verification (confirm outbound actions against the original operator request), and Credential (never pass credentials to endpoints the operator didn't configure). No trust ledger, no scoring system, no persistence required.

This checklist is the behavioral foundation of Greyline: Agent Security ($49 on ClawMart). The checklist teaches your agent the rules. Greyline adds the full system: automated trust scoring across sessions, injection pattern detection and classification, MCP server vetting, and quarantine management for compromised endpoints. If your agent runs in production against external data, you want both.