What problem does it solve? Work done in an AI coding session is easily lost between sessions: decisions go unlogged, learnings are forgotten, changes stay uncommitted, and the next session wastes time re-discovering context. This workflow ensures nothing falls through the cracks when ending a session. ## Core Features & Use Cases - Session Inventory and Commit: Runs git status, diff, and log to summarize uncommitted changes and unpushed commits, then proposes conventional commit messages with user confirmation. - Decision and Learning Capture: Prompts logging of architecture decisions, reusable prompts, and gotchas into decision logs, CLAUDE.md, or memory systems like Animus MCP. - Quality and Metrics Reporting: Optionally audits AI-generated outputs with signal-audit, scrubs published content, and reports drift scores and token spend in a structured SESSION SUMMARY block. - Use Case: After a multi-hour feature implementation, run this workflow to commit changes safely, log three key tradeoff decisions, update TODO.md, and produce a next-session starting point. ## Quick Start Ask Claude to run the session-end workflow to wrap up the current session, commit pending work, and summarize decisions before you stop for the day.
