← スキル一覧に戻る

compound
by DaveHudson
⭐ 0🍴 0📅 2026年1月11日
SKILL.md
name: compound description: Compound engineering workflow for self-improvement. Auto-loads when completing features, doing reviews, or capturing learnings into CLAUDE.md and rules.
Each unit of work should make future work easier. Capture learnings so knowledge compounds over time.
The Compound Loop
Plan (40%) → Work (20%) → Review (20%) → Compound (20%) → repeat
Traditional development accumulates debt. Compound engineering creates a learning loop.
After Completing Work
Ask these questions:
- What patterns were used or discovered?
- What decisions were made and why?
- What failures occurred? How to prevent them?
- What should be codified?
Where to Codify Learnings
| Learning Type | Location |
|---|---|
| Project conventions | CLAUDE.md |
| Domain rules | .claude/rules/<domain>.md |
| Reusable workflows | .claude/skills/ |
| Regression prevention | Test cases |
| Cross-project patterns | ~/.claude/CLAUDE.md |
What to Capture
In CLAUDE.md
- Project structure and conventions
- Tech stack decisions and rationale
- Common commands and workflows
- Team preferences
In Rules
- File-type specific patterns (e.g.,
paths: ["**/*.tsx"]) - Framework conventions (React, Convex, etc.)
- Code style beyond linting
In Skills
- Multi-step workflows
- Domain knowledge that auto-loads
- Reference materials
In Tests
- Bug regressions (prevent recurrence)
- Edge cases discovered
- Integration scenarios
Compound Prompts
After completing a feature or fix:
What did I learn that should be captured?
- [ ] New pattern → Add to CLAUDE.md or create rule
- [ ] Decision rationale → Document in CLAUDE.md
- [ ] Bug fix → Add regression test
- [ ] Workflow improvement → Update skill or create new one
Success Indicators
You're compounding well when:
- Similar features require less effort over time
- Bugs become one-time events (prevented by tests/rules)
- Code reviews surface fewer issues
- New team members onboard faster via CLAUDE.md
- Claude's suggestions align better with your codebase
Key Principles
- Document as you go - Don't defer knowledge capture
- Quality compounds - Good code enables good code
- Codify, don't memorize - Put it in files, not your head
- Prefer specific over general - Concrete examples > abstract rules
- Review periodically - Prune outdated learnings
スコア
総合スコア
50/100
リポジトリの品質指標に基づく評価
✓SKILL.md
SKILL.mdファイルが含まれている
+20
○LICENSE
ライセンスが設定されている
0/10
○説明文
100文字以上の説明がある
0/10
○人気
GitHub Stars 100以上
0/15
○最近の活動
3ヶ月以内に更新がある
0/10
○フォーク
10回以上フォークされている
0/5
✓Issue管理
オープンIssueが50未満
+5
✓言語
プログラミング言語が設定されている
+5
○タグ
1つ以上のタグが設定されている
0/5
レビュー
💬
レビュー機能は近日公開予定です