Analyzes Claude Code sessions to propose evidence-based skill improvements and capture persistent user preferences.
Reflect provides a structured framework for continuous improvement within the Claude Code environment by analyzing conversation signals such as user corrections, successes, and edge cases. It utilizes a validated workflow—including a specialized critic subagent—to suggest high-confidence updates to your local skills and memories, ensuring your AI assistant evolves and adapts to your specific workflow and project requirements over time.
Key Features
010 GitHub stars
02Automated session analysis for feedback and edge case detection
03Persistent memory management for cross-skill preferences
04Performance tracking and effectiveness metrics
05Skill validation against 12-factor agent principles
06Evidence-based improvement proposals with confidence ranking
Use Cases
01Optimizing custom skills based on real-world usage and failures
02Capturing and automating project-specific coding preferences
03Auditing existing skills for architectural compliance and reliability
