[BUG] System prompt hardcodes .claude path instead of resolving CLAUDE_CONFIG_DIR

Resolved 💬 3 comments Opened Apr 13, 2026 by jandudekcz Closed Apr 16, 2026

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?

When CLAUDE_CONFIG_DIR is set to a non-default path (e.g. .claude-oopix),
the auto-memory section of the system prompt still references the hardcoded
default .claude path. The model is told its memory lives at
C:\Users\<user>\.claude\projects\<slug>\memory\ even though the real folder
in use is C:\Users\<user>\.claude-oopix\projects\<slug>\memory\.

Memory read/write itself works correctly — only the prompt text is wrong,
which misleads the model when the user asks about memory location.

The system prompt contains a hardcoded .claude path regardless of the
CLAUDE_CONFIG_DIR value.

What Should Happen?

The path in the auto-memory system-prompt section should be templated from
CLAUDE_CONFIG_DIR, not hardcoded to .claude.

Error Messages/Logs

OS: Windows 11 Home 10.0.26200
Shell: bash (Git Bash)
Claude Code version: (run `claude --version` to fill in)
Launcher: VBS script that sets CLAUDE_CONFIG_DIR before spawning VS Code

Steps to Reproduce

  1. Set env var before launching:

set CLAUDE_CONFIG_DIR=C:\Users\me\.claude-oopix
claude

  1. Ask: "Where is your memory stored?"
  2. Model quotes the hardcoded .claude path from the system prompt

instead of .claude-oopix.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.92

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

Use case: running multiple Claude Code instances in parallel with different
CLAUDE_CONFIG_DIR values (one per subscription / project profile). Each
instance should report its own config path correctly.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗