self-improvement

To log errors and corrections for continuous improvement, you write entries to LEARNINGS.md and ERRORS.md inside a .learnings/ directory. This captures task outcomes to enable cross-session memory and recurring pattern extraction.

Cross-session learning promotion works by extracting recurring insights from logged corrections and writing them into CLAUDE.md, AGENTS.md, or Copilot instructions. This ensures broadly applicable learnings persist across future AI workflows.

The best way to capture recurring workflow patterns is to log errors and corrections in standard templates within the .learnings/ directory. When patterns recur, the skill promotes these insights directly into your Copilot instructions.

You need a SKILL.md file with frontmatter name and description, plus available assets, scripts, and references to support logging, extraction, and promotion. No external dependencies are required to start capturing learnings.

Yes, you can use standard templates and locations in the .learnings/ directory to log corrections for both single-session prompts and cross-session workflows. This standardization enables consistent learning extraction and continuous improvement.

Logged learnings may not persist if entries are not properly written to LEARNINGS.md and ERRORS.md, or if recurring patterns are not promoted to CLAUDE.md or AGENTS.md. Ensure standard templates and the .learnings/ directory are used correctly.