instruction-improver

You can update CLAUDE.md from user feedback by mining memex conversation history to identify recurring frustrations, corrections, and successes, then proposing and implementing explicit edits to your guidelines.

Turning conversation history into CLAUDE.md improvements involves reading existing rules, surfacing actionable feedback patterns from memex conversations, obtaining your explicit approval on findings, and implementing edits to chosen files.

Yes, you can apply CLAUDE.md improvements to both user-level and project-level files, allowing you to make explicit scope decisions and propose targeted changes for one or both configuration files as needed.

Extracting recurring mistakes and wins from memex conversations requires analyzing historical interactions to detect feedback patterns, which are then surfaced as actionable findings for your review and guideline updates.

Yes, you need to approve CLAUDE.md edits before implementation, as the workflow reads existing rules, presents identified feedback patterns as findings, and requires your explicit approval before applying any changes to the files.