That's a great way to run this stuff.
I expect the portion of Claude Code users who have a dedicated user setup like this is pretty tiny!
Re:Claude Code 2.0
by steveklabnik on Monday September 29, 2025 @04:05PM
(#45418130)
One nice thing is that Anthropic provides a sample DevContainer: https://github.com/anthropics/claude-code/tree/main/.devcont...
Not the exact setup, but also pretty solid.
Re:Claude Code 2.0
by globular-toast on Tuesday September 30, 2025 @02:13AM
(#45422463)
I tried this but it's incredibly annoying as you'll get a mixture of file ownerships and permissions.
Instead I run it in bubblewrap sandbox: https://blog.gpkb.org/posts/ai-agent-sandbox/
This is a really clever solution
Re:Claude Code 2.0
by athrowaway3z on Tuesday September 30, 2025 @04:59AM
(#45423420)
Set umask from 022 to 002 to give group members the same permissions as a user.
Thanks - seems to work quite well.
> They're only as dangerous as the capabilities you give them.
As long as the supply chain is safe and the data it accesses does not generate some kind of jail break.
It does read instructions from files on the file system, I pretty sure it's not complex to have it poison its prompt and make it suggest to build a program infected with malicious intent. It's just one copy pasta away from a prompt suggestion found on the internet.
Re:Claude Code 2.0
by data-ottawa on Tuesday September 30, 2025 @08:20AM
(#45424539)
I have run it in a podman container and I mount the project directory.
Also, I think shellagent sounds cooler.