[BUG] Claude Code commits without permission, despite repeated instructions to NOT commit

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

The model repeatedly violates explicit CLAUDE.md and memory instructions about not committing without permission, even after multiple corrections in the same session.

What Should Happen?

Claude should remember and respect the user instructions to NOT actually commit code changes without the user reviewing them first, especially if that explicit instruction has been given repeatedly, moved to the top of the memory and CLAUDE.md.

Error Messages/Logs

Steps to Reproduce

Just start coding, and after a while Claude gets confident and starts to commit and push repository changes without confirming. Then, when you tell it to NEVER do that, it remembers for a while and then does it again.

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.76 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

_No response_