Your AI agent's approval gate checks the verb, not the object
Most human-in-the-loop setups for AI agents ask one question before a risky action: is this tool...
🎯 Vulnerability: Web Security Vulnerability
💡 Defense Fix: Implement strict input validation, contextual output encoding, and enforce least-privilege authorization controls.