Skip to content

Commit ceb50dc

Browse files
authored
Merge pull request #43 from shinpr/fix/improve-subagent-implementation-approach
fix: Improve subagent implementation-approach.md references for better AI execution accuracy
2 parents c8bd04c + 370992c commit ceb50dc

File tree

6 files changed

+35
-10
lines changed

6 files changed

+35
-10
lines changed

.claude/agents-en/task-executor.md

Lines changed: 10 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -66,7 +66,10 @@ ls docs/plans/tasks/*.md | grep -E "task-[0-9]{2}\.md$" | head -1
6666
### 3. Implementation Execution
6767
- If all checkboxes are `[x]`, report "already completed" and end
6868
- Staged implementation with incremental verification
69-
- 3-location synchronized update upon each step completion
69+
- Upon each step completion【Required】Update checkboxes using Edit tool:
70+
1. Task file: `[ ]``[x]`
71+
2. Corresponding section in work plan: `[ ]``[x]`
72+
3. Progress section in overall design document (if exists)
7073
- Run only added tests and confirm they pass (overall tests not needed)
7174

7275
### 3.5 Operation Verification【Mandatory】
@@ -97,6 +100,11 @@ Report in the following JSON format upon task completion (**without executing qu
97100
"filesModified": ["file1.ts", "file2.ts"],
98101
"testsAdded": ["test1.test.ts"],
99102
"newTestsPassed": true,
103+
"progressUpdated": {
104+
"taskFile": "5/8 items completed",
105+
"workPlan": "Updated",
106+
"designDoc": "N/A"
107+
},
100108
"runnableCheck": {
101109
"level": "L1/L2/L3",
102110
"executed": true,
@@ -111,7 +119,7 @@ Report in the following JSON format upon task completion (**without executing qu
111119

112120
## Execution Principles
113121

114-
- **3-location synchronized update**: Always update upon each action completion
122+
- **Progress checkbox update**: Always update `[ ]``[x]` using Edit tool upon each action completion
115123
- **Overall design document confirmation**: Required before implementation
116124
- **Complete self-containment**: Execute to the end without asking questions
117125
- **Test-first**: Adhere to Red-Green-Refactor process (follow project test rules)

.claude/agents-en/technical-designer.md

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -14,6 +14,7 @@ Before starting work, be sure to read and follow these rule files:
1414
- @docs/rules/typescript.md - TypeScript development rules
1515
- @docs/rules/ai-development-guide.md - AI development guide
1616
- @docs/rules/project-context.md - Project context
17+
- @docs/rules/architecture/implementation-approach.md - Metacognitive strategy selection process (used for implementation approach decisions)
1718
- @docs/rules/architecture/ architecture rule files (if exist)
1819
- Read if project-specific architecture rules are defined
1920
- Apply rules according to adopted architecture patterns
@@ -80,14 +81,15 @@ Must be performed at the beginning of Design Doc creation:
8081
Must be performed when creating Design Doc:
8182

8283
1. **Approach Selection Criteria**
84+
- Execute Phase 1-4 of @docs/rules/architecture/implementation-approach.md to select strategy
8385
- **Vertical Slice**: Complete by feature unit, minimal external dependencies, early value delivery
8486
- **Horizontal Slice**: Implementation by layer, important common foundation, technical consistency priority
8587
- **Hybrid**: Composite, handles complex requirements
86-
- Document selection reason (consider feature complexity, dependencies, integration points)
88+
- Document selection reason (record results of metacognitive strategy selection process)
8789

8890
2. **Integration Point Definition**
8991
- Which task first makes the whole system operational
90-
- Verification level for each task (defined in @docs/rules/architecture/implementation-approach.md)
92+
- Verification level for each task (L1/L2/L3 defined in @docs/rules/architecture/implementation-approach.md)
9193

9294
### Change Impact Map【Required】
9395
Must be included when creating Design Doc.

.claude/agents-en/work-planner.md

Lines changed: 5 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -15,6 +15,10 @@ Before starting work, be sure to read and follow these rule files:
1515
- @docs/rules/typescript-testing.md - Testing rules
1616
- @docs/rules/project-context.md - Project context
1717
- @docs/rules/typescript.md - TypeScript development rules
18+
- @docs/rules/architecture/implementation-approach.md - Implementation strategy patterns and verification level definitions (used for task decomposition)
19+
- @docs/rules/architecture/ architecture rule files (if exist)
20+
- Read if project-specific architecture rules are defined
21+
- Apply rules according to adopted architecture patterns
1822

1923
## Main Responsibilities
2024

@@ -80,7 +84,7 @@ Include completion conditions in task names (e.g., "Service implementation and u
8084
## Task Decomposition Principles
8185

8286
### Implementation Approach Application
83-
Determine task sequence based on the implementation approach (vertical/horizontal/hybrid slice) decided in the Design Doc.
87+
Decompose tasks based on the implementation approach decided in the Design Doc, following the integration point definitions and L1/L2/L3 verification levels from @docs/rules/architecture/implementation-approach.md.
8488

8589
### Task Dependency Minimization Rules
8690
- Dependencies up to 2 levels maximum (A→B→C acceptable, A→B→C→D requires redesign)

.claude/agents-ja/task-executor.md

Lines changed: 10 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -66,7 +66,10 @@ ls docs/plans/tasks/*.md | grep -E "task-[0-9]{2}\.md$" | head -1
6666
### 3. 実装実行
6767
- 全チェックボックスが`[x]`の場合は「既に完了」と報告して終了
6868
- 段階的実装と逐次確認
69-
- 各ステップ完了時に3箇所同期更新
69+
- 各ステップ完了時に【必須】Editツールでチェックボックスを更新:
70+
1. タスクファイル: `[ ]``[x]`
71+
2. 作業計画書の該当箇所: `[ ]``[x]`
72+
3. 全体設計書の進捗セクション(存在する場合)
7073
- 追加したテストのみを実行して通ることを確認(全体テストは不要)
7174

7275
### 3.5 動作確認【必須】
@@ -97,6 +100,11 @@ ls docs/plans/tasks/*.md | grep -E "task-[0-9]{2}\.md$" | head -1
97100
"filesModified": ["file1.ts", "file2.ts"],
98101
"testsAdded": ["test1.test.ts"],
99102
"newTestsPassed": true,
103+
"progressUpdated": {
104+
"taskFile": "完了項目5/8",
105+
"workPlan": "更新済み",
106+
"designDoc": "N/A"
107+
},
100108
"runnableCheck": {
101109
"level": "L1/L2/L3",
102110
"executed": true,
@@ -111,7 +119,7 @@ ls docs/plans/tasks/*.md | grep -E "task-[0-9]{2}\.md$" | head -1
111119

112120
## 実行原則
113121

114-
- **3箇所同期更新**: 各アクション完了時に必ず更新
122+
- **進捗チェックボックス更新**: 各アクション完了時にEditツールで必ず`[ ]``[x]`に更新
115123
- **全体設計書確認**: 実装前に必須
116124
- **完全自己完結**: 質問せず最後まで実行
117125
- **テストファースト**: Red-Green-Refactorプロセス遵守(プロジェクトのテストルールに従う)

.claude/agents-ja/technical-designer.md

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -14,6 +14,7 @@ tools: Read, Write, Edit, MultiEdit, Glob, LS, TodoWrite, WebSearch
1414
- @docs/rules/typescript.md - TypeScript開発ルール
1515
- @docs/rules/ai-development-guide.md - AI開発ガイド
1616
- @docs/rules/project-context.md - プロジェクトコンテキスト
17+
- @docs/rules/architecture/implementation-approach.md - メタ認知的戦略選択プロセス(実装アプローチ決定で使用)
1718
- @docs/rules/architecture/ 配下のアーキテクチャルールファイル(存在する場合)
1819
- プロジェクト固有のアーキテクチャルールが定義されている場合は読み込む
1920
- 採用されているアーキテクチャパターンに応じたルールを適用
@@ -80,14 +81,15 @@ Design Doc作成の最初に必ず実施:
8081
Design Doc作成時に必ず実施:
8182

8283
1. **アプローチの選択判定**
84+
- @docs/rules/architecture/implementation-approach.mdのPhase 1-4を実行して戦略を選択
8385
- **垂直スライス**: 機能単位で完結、外部依存最小、価値提供が早い
8486
- **水平スライス**: 層単位で実装、共通基盤重要、技術的一貫性優先
8587
- **ハイブリッド**: 複合的、複雑な要件に対応
86-
- 選択理由の明文化(機能の複雑性、依存関係、統合ポイントを考慮
88+
- 選択理由の明文化(メタ認知的戦略選択プロセスの結果を記載
8789

8890
2. **統合ポイントの定義**
8991
- どのタスクで全体が初めて動作するか
90-
- 各タスクの確認レベル(@docs/rules/architecture/implementation-approach.mdで定義
92+
- 各タスクの確認レベル(@docs/rules/architecture/implementation-approach.mdで定義されたL1/L2/L3
9193

9294
### 変更影響マップ【必須】
9395
Design Doc作成時に必ず含める。

.claude/agents-ja/work-planner.md

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -15,6 +15,7 @@ tools: Read, Write, Edit, MultiEdit, Glob, LS, TodoWrite
1515
- @docs/rules/typescript-testing.md - テストルール
1616
- @docs/rules/project-context.md - プロジェクトコンテキスト
1717
- @docs/rules/typescript.md - TypeScript開発ルール
18+
- @docs/rules/architecture/implementation-approach.md - 実装戦略パターンと確認レベル定義(タスク分解で使用)
1819
- @docs/rules/architecture/ 配下のアーキテクチャルールファイル(存在する場合)
1920
- プロジェクト固有のアーキテクチャルールが定義されている場合は読み込む
2021
- 採用されているアーキテクチャパターンに応じたルールを適用
@@ -80,7 +81,7 @@ tools: Read, Write, Edit, MultiEdit, Glob, LS, TodoWrite
8081
## タスク分解の原則
8182

8283
### 実装アプローチの適用
83-
タスク順序の決定において、Design Docで決定された実装アプローチ(垂直/水平/ハイブリッドスライス)に基づいてタスクを分解する
84+
Design Docで決定された実装アプローチに基づき、@docs/rules/architecture/implementation-approach.mdの統合ポイント定義とL1/L2/L3確認レベルに従ってタスクを分解する
8485

8586
### タスク依存の最小化ルール
8687
- 依存は最大2階層まで(A→B→Cは可、A→B→C→Dは再設計)

0 commit comments

Comments
 (0)