The Coding Sessions you should be learning from

In this walkthrough I cover:

• Why sessions lose their value — token waste, vague prompts, and suboptimal agent routing are patterns that repeat across sessions but rarely get addressed

• A two-phase analysis loop — a local transcript parser compresses the session log by 90–95%, then Claude analyzes the digest to extract learnings and scores

• How it hooks into Claude Code — the stop lifecycle event fires automatically at session end, triggering the analysis with zero manual effort

• What gets persisted — a rolling SKILL.md of top findings plus per-session detail logs, with promotion and decay logic for recurring issues

• What’s configurable — file size thresholds, turn limits, session retention count, promotion windows — all tunable per repo

• Where this goes next — session score dashboards, team-shared learnings, pre-session coaching hooks

The repo includes an install script — one curl command drops the hook and skill into your .claude directory and it starts working from the next session.

Would love feedback on what you’d add to the roadmap or how you’d extend the feedback loop for your own workflow.

The author is building Auron — an AI-powered voice and conversation intelligence platform that captures and enriches organizational knowledge from meetings, calls, and conversations. Auron turns every interaction into structured signal that teams can act on.