Language: English | 日本語
An Agent Skill that extracts reusable patterns from Claude Code sessions, self-evaluates their quality, and determines the right save location (Global vs Project) before saving.
cp -r skills/learn-eval ~/.claude/skills/learn-evalComing soon.
The skill follows a 7-step process:
- Review the session for extractable patterns
- Identify the most valuable/reusable insight
- Determine save location — Global (
~/.claude/skills/learned/) for cross-project patterns, Project (.claude/skills/learned/) for project-specific knowledge - Draft the skill file using a standardized format
- Quality gate — run checklist + holistic judgment
- Confirm with the user (verdict-specific confirmation flow)
- Save or absorb into the determined location
Every candidate pattern goes through a two-part evaluation:
- Grep
~/.claude/skills/for keyword overlap — confirmed no content duplication - Check MEMORY.md (project + global) for duplication
- Consider whether appending to an existing skill is sufficient
- Confirm the pattern is reusable, not a one-off fix
Based on the checklist results, one of four verdicts is issued:
| Verdict | Meaning | Next Action |
|---|---|---|
| Save | Unique, specific, properly scoped | Proceed to save |
| Improve then Save | Valuable but needs refinement | List improvements, revise, re-evaluate (once) |
| Absorb into [X] | Should be appended to an existing skill | Show target + diff, then save |
| Drop | Trivial, redundant, or too abstract | Explain reason and stop |
- Error Resolution Patterns — root cause + fix + reusability
- Debugging Techniques — non-obvious steps, tool combinations
- Workarounds — library quirks, API limitations, version-specific fixes
- Project-Specific Patterns — conventions, architecture decisions, integration patterns
- Trivial fixes (typos, simple syntax errors)
- One-time issues (specific API outages, temporary version bugs)
- Patterns that can be found by searching the error message
- Standard documentation-level knowledge
skills/learn-eval/knowledge-placement-decision.md— A decision tree for choosing between Memory, existing skill append, and new skill creation. Prevents skill sprawl and knowledge burial.