Is NanoClaw safer than OpenClaw?

Yes, by design. OpenClaw often runs agents in a single shared process, meaning one compromised agent could potentially access all your data. NanoClaw uses OS-level container isolation (Apple Containers or Docker), preventing agents from accessing any file or directory you have not explicitly authorized.

How do I prevent an AI agent from deleting my files?

NanoClaw uses hard boundaries. Instead of relying on instructions like "do not delete files," it runs the agent in a sandbox where it cannot see or touch system files. The agent only has read and write access to the specific folders you mount.

What is the difference between an AI "Skill" and a "Plugin"?

Traditional plugins are often always-on and can be over-permissioned, which increases security risk. A NanoClaw Skill is modular and is used to modify your local code only when needed. This reduces permission creep because you add only the code required for a task rather than a large library of unused features.

Can NanoClaw agents work together on one task?

Yes. This is called an Agent Swarm. For complex workflows such as researching a topic and writing a summary, NanoClaw can orchestrate multiple specialized agents to collaborate, improving accuracy and reducing errors.

Do I need to be a developer to add new "Skills" to NanoClaw?

Not necessarily. You can describe what you want in natural language, for example "connect my assistant to my Todoist account." NanoClaw uses its skill-building logic to update its own local code for you.