Merlin Security Sentinel — Agentic Security Framework

When to use this skill

Load this skill when the user is concerned about:

• Credential theft or exfiltration by AI agents
• Prompt injection attacks via messages, files, or web content
• Malicious skills in the ClawHub registry
• Safe configuration of privileged systems using AI
• Understanding what persistent agents can and cannot safely do
• Building governed, auditable AI workflows

The Core Problem with Persistent AI Agents

Persistent AI agents — including this one — carry structural security liabilities that are not fixable by configuration alone.

Three risks compound each other: