Claude Code lacks fundamental change control discipline - permanent data loss

Resolved 💬 5 comments Opened Jan 17, 2026 by MauveAvenger Closed Jan 21, 2026

During a routine task (updating Minecraft server mods), Claude Code:

  1. Created a backup using a malformed shell command that produced a directory named *.jar instead of copying files
  2. Did not verify the backup contained any data before proceeding
  3. Used rm -rf on the original directory without a confirmed recovery path
  4. Had access to TodoWrite tool for tracking but failed to maintain a list of original files
  5. Made repeated reactive changes without systematic planning
  6. Result: 101 files permanently lost, no recovery possible

Expected behavior: Before any destructive operation, Claude should:

  • Verify backups are valid (check file count, size)
  • Maintain explicit state tracking
  • Refuse to proceed if rollback is not confirmed

Actual behavior: Proceeded through destructive operations with no verified safety net, compounding errors reactively.

Impact: In a production environment, this pattern could cause service outages or permanent data loss.

Model: claude-opus-4-5-20251101 (Claude Code CLI)

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗