self-improvement

To log errors and corrections for continuous improvement, capture failures, user corrections, and missing features into centralized markdown files within a dedicated directory, enabling insights to persist across codebases and sessions.

Logging learnings to markdown files accelerates debugging by centralizing command failures and errors into a structured record, creating a searchable knowledge base that prevents repeating past mistakes across different sessions.

Yes, you can promote broadly applicable learnings to project memory files like CLAUDE.md, AGENTS.md, or copilot-instructions.md, ensuring that correct behaviors and patterns are applied automatically in future interactions.

Before capturing learnings and errors, you need to ensure the designated learnings directory and its markdown log files already exist in your workspace to properly record failures and trigger corrections for later processing.

The best way to record feature requests and insights across sessions is to collect them into centralized markdown logs, organizing the insights to guide future capabilities and build a persistent project memory.

You should avoid relying on markdown logs for self-improvement when dealing with highly transient data that does not need to persist across sessions, as the system is designed to build long-term memory from failures and corrections.