Comment by philipp-gayret

My workflow is somewhat similar to yours. I also much love --dangerously-skip-permissions, as root! I even like to do it from multiple Claude Code instances in parallel when I have parallel ideas that can be worked out.

Maybe my wrapper project is interesting for you? https://github.com/release-engineers/agent-sandbox It's to keep Claude Code containerized with a copy of the workspace and a firewall/proxy so it can only access certain sites. With my workflow I don't really risk much, and the "output" is a .patch file I can inspect before I git apply it.