[BUG] # [insert memory] command overwrites CLAUDE.md instead of appending to it
Resolved 💬 4 comments Opened Jun 6, 2025 by jm7610 Closed Dec 10, 2025
Environment
- Platform (select one):
- [ ] Anthropic API
- [ ] AWS Bedrock
- [ ] Google Vertex AI
- [X] Other: Claude Code
- Claude CLI version: v1.0.16
- Operating System: macOS 15.5
- Terminal: Terminal App
## Bug Description
The # [insert memory] terminal command completely overwrote the CLAUDE.md file instead of appending new content to it, resulting in loss of all existing project documentation and guidance.
## Steps to Reproduce
- Have an existing CLAUDE.md file with comprehensive project documentation (multiple sections,~800+ lines)
- Use the
# [insert memory]command in terminal to add a new memory/guidance - Check the CLAUDE.md file contents
## Expected Behavior
The # [insert memory] command should:
- Append the new memory/guidance to the existing CLAUDE.md file
- Preserve all existing content
## Actual Behavior
The # [insert memory] command:
- Completely overwrote the entire CLAUDE.md file
- Replaced all existing content with only the new memory
- Resulted in a file with just a few lines instead of the original comprehensive documentation
Example of what happened:
- Original CLAUDE.md: ~800+ lines with sections for pipeline architecture, development principles, troubleshooting, etc.
- After
# [insert memory]: Only 8 lines remaining with just the header and single new memory
## Additional Context
This is a critical issue because:
- CLAUDE.md files often contain essential project-specific instructions that guide Claude Code's behavior
- The documentation may include architectural decisions, coding standards, and complex workflows that are difficult to recreate
- Users may not realize their documentation has been overwritten until they notice unexpected behavior
Suggested improvements:
- Make
# [insert memory]append rather than overwrite - Add a confirmation prompt if overwriting will occur
- Create automatic backups before modifying CLAUDE.md
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗