self-improvement

To track errors and corrections across development sessions, you log failures and user corrections into standardized Markdown files under a .learnings/ directory. Entries include consistent IDs, priorities, status fields, and structured metadata for searchable, actionable continuous improvement.

Logging knowledge gaps and feature requests in markdown involves recording details in structured entries within .learnings/ files. Standardized formats capture summaries, suggested actions, priorities, and area tags to make project knowledge searchable and actionable.

Creating a continuous improvement workflow for failed commands involves running provided hook scripts that prompt evaluation after tasks. These deterministic scripts encourage error logging after external tool failures and capture the recommended fix.

You should promote recurring learnings into permanent project memory when better approaches are discovered repeatedly across sessions. The skill provides guidance to decide when to promote entries to memory files or extract them into a standalone SKILL.md.

You can use structured markdown for error tracking without external dependencies because the skill requires none. It relies solely on saving standardized Markdown entries with consistent IDs and metadata fields under the .learnings/ directory.

A limitation when extracting skills from outdated knowledge realizations is deciding the correct promotion path. You must manually evaluate whether to append the realization to project memory or extract it into a standalone SKILL.md for reusable agent workflows.