napkin

Track mistakes across sessions by logging corrections and patterns into a per-repo markdown file. The napkin approach reads and updates this file before and during work to maintain a persistent working memory of lessons learned.

A repository-level working memory system is a persistent log that records mistakes, corrections, and patterns. It is read and updated before and during work sessions to inform future development tasks and reduce repeated errors across the project.

Start logging session mistakes by creating a markdown file at .claude/napkin.md in your repository. Read and update this file at the start of every session and capture key insights immediately as you work using the provided template.

No, this session logging approach requires zero external dependencies or components. It operates as a lightweight self-improvement loop using a minimal markdown template that teams can tailor to their specific project domains and workflows.

The best way to maintain a learning memory for coding errors is applying a consistent logging loop. Use health checks and evaluation guidance to ensure new insights are captured immediately and the log is reviewed before starting new work.

You should not use a napkin file for tracking development patterns if your workflow already enforces structured documentation or automated linting that covers error tracking. It serves as a lightweight, adaptable loop rather than a rigid compliance system.