[stavros]

I made a secure one:

https://github.com/skorokithakis/stavrobot

Everything runs in containers (I run it on a server along with everything else), plugins have a permission system so eg the AI can read emails but not delete or send, etc.

I really like it, I run it as my main agent and it has been extremely helpful.

[croes]

Part of the usefulness is based on the same thing that makes it so dangerous.

If it can only read but not act, it’s safer but less useful.

[stavros]

I can't restrict OpenClaw if I don't need the extra capabilities. I can restrict this.

[croes]

You restrict OpenClaw by not providing it certain credentials.

[stavros]

Again, with my design you can give it fine-grained access to parts of services, which OpenClaw itself cannot do. This is just a fact.