session-reflection

Record team session learnings by running a structured reflection workflow at the end of each session to append categorized lessons, mistakes, and patterns to the project's .claude.md file. This creates a persistent memory that enables self-evolution across multiple agent or team work sessions.

A structured post-mortem reflection process classifies session evidence into what worked, what didn't, and what to change. It derives actionable rules from these categories and emits them to project memory, ensuring that human instruction pitfalls and agent mistakes translate into continuous process improvements.

Capture agent mistakes and human instruction pitfalls by executing an end-of-session retrospective workflow that records errors and valuable methods. The workflow appends these findings to .claude.md, storing classified evidence and derived rules for persistent team memory and future avoidance.

Session reflection works for non-software collaboration tasks because it applies to any team or agent-session requiring post-mortem learning. The structured reflection process captures valuable methods and lessons across general collaboration tasks, not just software development, by classifying evidence universally.

You can save retrospective rules to universal memory in addition to project memory. The reflection workflow emits derived rules to the project's .claude.md by default, while providing an optional universal memory output for broader cross-project learning and pattern retention.

You need to run an end-of-session retrospective at the close of every team or work session to document lessons and patterns. Running it consistently ensures that mistakes, valuable methods, and process changes are continuously captured for persistent memory and self-evolution.