# Iterative Refinement

> Build→Test→Feedback→Refineの反復改善ワークフロー。ツール開発時のフィードバックループとOSS検証を支援。

- Skill: `nwiizo/iterative-refinement` (Agent Skill)
- Install (CLI): `npx skillmds@latest add nwiizo/iterative-refinement`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nwiizo/iterative-refinement/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: nwiizo (https://skillmd.com/u/nwiizo)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/nwiizo/iterative-refinement

---


# Iterative Refinement

Build → Test → Feedback → Refine → Repeat

## Phases

1. **Initial**: コア機能をまず動かす。最適化は後
2. **Internal Test**: 自分のコードベースでテスト。違和感をメモ
3. **User Feedback**: 3つの質問で確認
   - ツールとしてどうか？
   - 使いやすいか？
   - ユーザーとして使いたいか？
4. **Refinement**: 優先度順に修正
   1. Breaking issues → 2. Noise reduction → 3. Clarity → 4. Polish
5. **External Validation**: OSS プロジェクトで検証 (→ validate-on-oss)

## Common Feedback → Action

| Feedback | Action |
|----------|--------|
| Too much noise | Strict mode, hide low-priority |
| Tables broken | Switch to bullet points |
| Need language support | Add --jp flag |
| Not actionable | Add specific suggestions |

## Learnings Template

```markdown
### What Didn't Work
- {feature}: {why}

### What Worked
- {feature}: {why}

### Key Insight
{one sentence}
```

## Example: cargo-coupling
1. Initial: 全 issue 表示 → 2. Feedback: "60件はノイズ" → 3. Strict mode 追加 → 4. 3件の actionable issue のみ表示
Learning: "Default to actionable, opt-in to verbose"