Setup

---
description: Interactive setup for EDAF v1.0 Self-Adapting System / EDAF v1.0 自己適応型システムのインタラクティブセットアップ
---

# EDAF v1.0 - Interactive Setup / インタラクティブセットアップ

Welcome to EDAF (Evaluator-Driven Agent Flow) v1.0!
EDAF (Evaluator-Driven Agent Flow) v1.0へようこそ！

This setup wizard will help you configure the self-adapting workers and evaluators for your project.
このセットアップウィザードは、プロジェクトの自己適応型ワーカーとエバリュエーターを設定するお手伝いをします。

---

## Step 0: Language Preferences / ステップ0: 言語設定

**IMPORTANT / 重要**: Please select your language preference first. This will affect how Claude Code responds to you and generates documentation.
まず言語設定を選択してください。これにより、Claude Codeがどのように応答し、ドキュメントを生成するかが決まります。

**Action / アクション**: Use AskUserQuestion to select language preference:

```typescript
const fs = require('fs')
const path = require('path')

const langResponse = await AskUserQuestion({
  questions: [
    {
      question: "Select your language preference for EDAF / EDAFの言語設定を選択してください",
      header: "Language / 言語",
      multiSelect: false,
      options: [
        {
          label: "Option 1: EN docs + EN output",
          description: "Documentation in English, Terminal output in English. Best for English-speaking teams. / 英語ドキュメント、英語出力。英語圏チームに最適。"
        },
        {
          label: "Option 2: JA docs + JA output / 日本語ドキュメント + 日本語出力",
          description: "Documentation in Japanese, Terminal output in Japanese. Best for Japanese teams. / 日本語ドキュメント、日本語出力。日本語チームに最適。"
        },
        {
          label: "Option 3: EN docs + JA output / 英語ドキュメント + 日本語出力",
          description: "Documentation in English, Terminal output in Japanese. Best for learning English while working in Japanese. / 英語ドキュメント、日本語出力。英語学習しながら日本語で作業。"
        },
        {
          label: "Option 4: EN docs + JA output + JA translation / 英語ドキュメント + 日本語出力 + 日本語翻訳",
          description: "Documentation in English with Japanese translation saved separately. Terminal output in Japanese. Best for bilingual teams. / 英語ドキュメントと日本語翻訳を両方保存。ターミナル出力は日本語。バイリンガルチームに最適。"
        }
      ]
    }
  ]
})

// Parse the selected option
const selected = langResponse.answers['0']
let docLang = 'en'
let termLang = 'en'
let dualDocs = false

if (selected.includes('Option 1')) {
  docLang = 'en'
  termLang = 'en'
  dualDocs = false
} else if (selected.includes('Option 2')) {
  docLang = 'ja'
  termLang = 'ja'
  dualDocs = false
} else if (selected.includes('Option 3')) {
  docLang = 'en'
  termLang = 'ja'
  dualDocs = false
} else if (selected.includes('Option 4')) {
  docLang = 'en'
  termLang = 'ja'
  dualDocs = true
}

console.log('\n✅ Language preference set / 言語設定が完了しました:')
console.log('   Documentation / ドキュメント:', docLang === 'en' ? 'English / 英語' : 'Japanese / 日本語')
console.log('   Terminal Output / ターミナル出力:', termLang === 'en' ? 'English / 英語' : 'Japanese / 日本語')
if (dualDocs) {
  console.log('   Dual Language Docs / 二言語ドキュメント: Enabled (EN + JA) / 有効（英語 + 日本語）')
}

Next / 次: Save configuration and generate CLAUDE.md

// NOTE: Configuration will be saved after Docker detection in Step 2
// Docker検出後（Step 2）に設定を保存します
console.log('\n📝 Language preferences will be saved after environment detection')
console.log('📝 言語設定は環境検出後に保存されます')

// Generate CLAUDE.md
let claudeMd = `---
description: EDAF v1.0 Configuration and Language Preferences
---

# EDAF v1.0 - Claude Code Configuration

## Language Preferences

This file is auto-generated by \`/setup\` command.
Do not edit manually - run \`/setup\` again to change preferences.

**Current Settings:**

- **Documentation Language**: ${docLang === 'en' ? 'English' : 'Japanese'}
- **Terminal Output Language**: ${termLang === 'en' ? 'English' : 'Japanese'}
- **Save Dual Language Docs**: ${dualDocs ? 'Yes (EN + JA)' : 'No'}

---

## EDAF 4-Phase Gate System - IMPORTANT

**When the user requests to implement a feature using "エージェントフロー" (agent flow) or "EDAF", you MUST follow this exact workflow:**

### Phase 1: Design Gate
1. Launch \`designer\` agent via Task tool
2. Designer creates design document in \`docs/designs/{feature-slug}.md\`
3. Launch ALL 7 design evaluators in parallel via Task tool:
   - design-consistency-evaluator
   - design-extensibility-evaluator
   - design-goal-alignment-evaluator
   - design-maintainability-evaluator
   - design-observability-evaluator
   - design-reliability-evaluator
   - design-reusability-evaluator
4. Review evaluation results
5. If evaluators request changes, ask designer to revise
6. Repeat until all evaluators approve (≥ 7.0/10.0)

### Phase 2: Planning Gate
1. Launch \`planner\` agent via Task tool
2. Planner creates task plan in \`docs/plans/{feature-slug}-tasks.md\`
3. Launch ALL 7 planner evaluators in parallel via Task tool:
   - planner-clarity-evaluator
   - planner-deliverable-structure-evaluator
   - planner-dependency-evaluator
   - planner-goal-alignment-evaluator
   - planner-granularity-evaluator
   - planner-responsibility-alignment-evaluator
   - planner-reusability-evaluator
4. Review evaluation results
5. If evaluators request changes, ask planner to revise
6. Repeat until all evaluators approve (≥ 7.0/10.0)

### Phase 2.5: Implementation
1. Launch appropriate worker agents via Task tool based on task plan:
   - database-worker-v1-self-adapting (for database models)
   - backend-worker-v1-self-adapting (for backend logic)
   - frontend-worker-v1-self-adapting (for UI components)
   - test-worker-v1-self-adapting (for tests)
2. Workers implement code according to task plan

### Phase 3: Code Review Gate
1. Launch ALL 7 code evaluators in parallel via Task tool:
   - code-quality-evaluator-v1-self-adapting
   - code-testing-evaluator-v1-self-adapting
   - code-security-evaluator-v1-self-adapting
   - code-documentation-evaluator-v1-self-adapting
   - code-maintainability-evaluator-v1-self-adapting
   - code-performance-evaluator-v1-self-adapting
   - code-implementation-alignment-evaluator-v1-self-adapting
2. Review evaluation results
3. **If fr