claude-reflect is a self-learning enhancement system for Claude Code that captures user corrections, positive feedback, and preferences during interactive coding sessions and turns them into persistent knowledge that improves future responses. It watches what you correct Claude about — such as preferring a particular model, style, or workflow — and automatically queues those learnings, then lets you review and sync them back into configuration files like CLAUDE.md and agent definitions so Claude remembers them across sessions. Over time, this creates a personalized memory that helps the AI align more closely with your conventions, avoiding repeated misunderstandings and reducing friction in long-running or recurring tasks. In addition to capturing corrections, claude-reflect analyzes session history to identify repeated patterns and propose reusable commands or workflows, which can be converted into skills that accelerate productivity.
Features
- Captures corrections and feedback automatically
- Syncs learnings into Claude Code config files
- Builds persistent memory for future sessions
- Identifies repeating patterns for reusable skills
- Suggests new commands based on history
- Improves customization over time