[BUG] Claude stops using claude.md information

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?

Hey folks,

when having a longer session lets say 100k token, claude ignores rules we defined in claude.md - it slides out of context window. This is hardly annoying and should be solved.

In my cases I would prefer some settings like: "How often claude has to look into claude.md" or an "indicator telling a timestamp of the last time claude read the whole claude.md file".

For autonomous coding claude has to been forced to use a code of conduct like the claude.md file - without any excuses.

Stay tuned!

What Should Happen?

Claude has to lookup more often into claude.md and has to be forced doing it!

Error Messages/Logs

none

Steps to Reproduce

1. Create claude.md "Never ever use unicode emotes in python command line output"
2. Start chat into complex python session
3. After 1 hour of coding let claude to some python stuff with outputs where it possible for claude "to pimp" something by itself.
4. Nearly everytime I have emotes on my command line output like a damn comic book after a while.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.0.57

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_