[BUG]
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?
PROBLEM: Claude prompts to read global CLAUDE.md
CONTEXT: I regularly update ~/.claude/CLAUDE.md to adjust Claude behavior. In fact, in that CLAUDE.md, I define a "sync" command that tells Claude to read the global CLAUDE.md. The command works and Claude tries to read global CLAUDE.md but it prompts me with Read(~/.claude/CLAUDE.md) 1...2...3...
The prompt makes no sense and is quite annoying since I told claude to read its own CLAUDE.md file and it can always read it so why on earth is it asking me for permission to read the global CLAUDE.md. This gets especially annoying with multiple Claude Code windows running and each one is asking for permission to read the global CLAUDE.md. There is no reason to prevent Claude Code from reading its own CLAUDE.md because the design intent of that file is to have Claude actually read it.
What Should Happen?
Claude Code should never prompt to read its own CLAUDE.md files that it has to read by design
Error Messages/Logs
Steps to Reproduce
Step 1: Create ~/.claude/CLAUDE.md as follows:
BEGIN CLAUDE.md
sync
Developer issues sync command with the expectation that:
- Claude IMMEDIATELY re-reads project CLAUDE.md
- Claude IMMEDIATELY re-reads ~/.claude/CLAUDE.md
END CLAUDE.md
Step 2: Create empty folder foo
cd foo
claude
sync
sync
The first sync does not prompt (yay!). The second sync prompts (AUUUUGH!)
Claude Model
Sonnet (default)
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.0.30
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
I'm just using Claude Code with Pro subscription. I have no idea what API Platform Claude Code uses since I never specified any Platform. Perhaps the creepiest thing about this bug is that Claude INSISTS it never prompted me. AI gaslights human.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗