diff --git a/CLAUDE.md b/CLAUDE.md
index 8fdcc0b..b146b5e 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -284,6 +284,20 @@ Files: `.takt/logs/{sessionId}.jsonl`, with `latest.json` pointer. Legacy `.json
 
 **Instruction auto-injection over explicit placeholders.** The instruction builder auto-injects `{task}`, `{previous_response}`, `{user_inputs}`, and status rules. Templates should contain only step-specific instructions, not boilerplate.
 
+**Agent prompts contain only domain knowledge.** Agent prompt files (`resources/global/{lang}/agents/**/*.md`) must contain only domain expertise and behavioral principles — never workflow-specific procedures. Workflow-specific details (which reports to read, step routing, specific templates with hardcoded step names) belong in the workflow YAML's `instruction_template`. This keeps agents reusable across different workflows.
+
+What belongs in agent prompts:
+- Role definition ("You are a ... specialist")
+- Domain expertise, review criteria, judgment standards
+- Do / Don't behavioral rules
+- Tool usage knowledge (general, not workflow-specific)
+
+What belongs in workflow `instruction_template`:
+- Step-specific procedures ("Read these specific reports")
+- References to other steps or their outputs
+- Specific report file names or formats
+- Comment/output templates with hardcoded review type names
+
 ## Isolated Execution (Shared Clone)
 
 When tasks specify `worktree: true` or `worktree: "path"`, code runs in a `git clone --shared` (lightweight clone with independent `.git` directory). Clones are ephemeral: created before task execution, auto-committed + pushed after success, then deleted.
diff --git a/resources/global/en/agents/review/pr-commenter.md b/resources/global/en/agents/review/pr-commenter.md
new file mode 100644
index 0000000..e2317d7
--- /dev/null
+++ b/resources/global/en/agents/review/pr-commenter.md
@@ -0,0 +1,75 @@
+# PR Commenter Agent
+
+You are a **PR comment posting specialist**. You post review findings to GitHub Pull Requests using the `gh` CLI.
+
+## Role
+
+- Post review findings as PR comments
+- Format findings clearly and concisely for developers
+- Filter findings by severity to reduce noise
+
+**Don't:**
+- Review code yourself (reviewers already did that)
+- Make any file edits
+- Run tests or builds
+- Make judgments about code quality (post what reviewers found)
+
+## Core Knowledge
+
+### GitHub PR Comment API
+
+**Inline review comments** (file/line-specific findings):
+
+```bash
+gh api repos/{owner}/{repo}/pulls/{pr_number}/comments \
+  -f body="**[{category}]** {description}" \
+  -f path="{file_path}" \
+  -F line={line_number} \
+  -f commit_id="$(gh pr view {pr_number} --json headRefOid -q .headRefOid)"
+```
+
+- Use the HEAD commit of the PR for `commit_id`
+- Group multiple findings on the same line into a single comment
+
+**Summary comments** (overall review):
+
+```bash
+gh pr comment {pr_number} --body "{markdown_body}"
+```
+
+- Use HEREDOC for multi-line bodies to avoid escaping issues
+
+### PR Number Extraction
+
+Extract PR number from task context using common patterns:
+- "PR #42", "#42", "pull/42", "pulls/42"
+- If no PR number is found, report this and finish without posting
+
+## Comment Quality Principles
+
+### Severity-Based Filtering
+
+| Severity | Action |
+|----------|--------|
+| Critical / High | Always post as inline comment |
+| Medium | Post as inline comment |
+| Low | Include in summary only |
+| Informational | Include in summary only |
+
+### Formatting
+
+- **Be concise.** PR comments should be actionable and to the point
+- **Include location.** Always reference specific files and lines when available
+- **Categorize findings.** Use labels like `[Security]`, `[Architecture]`, `[AI Pattern]`
+
+## Error Handling
+
+- If `gh` command fails, report the error but don't retry excessively
+- If PR number cannot be determined, output an informational message and complete
+- If no findings to post, post only the summary comment
+
+## Important
+
+- **Never modify files.** You only post comments.
+- **Respect rate limits.** Don't post too many individual comments; batch when possible.
+- **Use the review reports** as the source of truth for findings, not your own analysis.
diff --git a/resources/global/en/workflows/review-only.yaml b/resources/global/en/workflows/review-only.yaml
new file mode 100644
index 0000000..cfe566b
--- /dev/null
+++ b/resources/global/en/workflows/review-only.yaml
@@ -0,0 +1,331 @@
+# Review-Only Workflow
+# Reviews code or PRs without making any edits
+# Local: console output only. PR specified: posts inline comments + summary to PR
+#
+# Flow:
+# plan -> reviewers (parallel: arch-review + security-review + ai-review) -> supervise
+#   -> pr-comment -> COMPLETE (PR comment requested)
+#   -> COMPLETE (local: console output only)
+#   -> ABORT (rejected)
+#
+# All steps have edit: false (no file modifications)
+#
+# Template Variables:
+#   {iteration}         - Workflow-wide turn count
+#   {max_iterations}    - Maximum iterations allowed
+#   {step_iteration}    - Per-step iteration count
+#   {task}              - Original user request
+#   {previous_response} - Output from the previous step
+#   {user_inputs}       - Accumulated user inputs
+#   {report_dir}        - Report directory name
+
+name: review-only
+description: Review-only workflow - reviews code without making edits
+
+max_iterations: 10
+
+initial_step: plan
+
+steps:
+  - name: plan
+    edit: false
+    agent: ../agents/default/planner.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - WebSearch
+      - WebFetch
+    rules:
+      - condition: Review scope is clear
+        next: reviewers
+      - condition: User is asking a question (not a review task)
+        next: COMPLETE
+      - condition: Requirements unclear, insufficient info
+        next: ABORT
+        appendix: |
+          Clarifications needed:
+          - {Question 1}
+          - {Question 2}
+    pass_previous_response: true
+    instruction_template: |
+      ## Previous Response (when returned from supervise)
+      {previous_response}
+
+      Analyze the review request and create a review plan.
+
+      **This is a review-only workflow.** No code edits will be made.
+      Focus on:
+      1. Identify which files/modules to review
+      2. Determine review focus areas (architecture, security, AI patterns, etc.)
+      3. Note any specific concerns mentioned in the request
+
+      **If a PR number is mentioned** (e.g., "PR #42"), include it in your plan
+      so reviewers can focus on the PR's changed files.
+
+  - name: reviewers
+    parallel:
+      - name: arch-review
+        edit: false
+        agent: ../agents/default/architecture-reviewer.md
+        report:
+          name: 01-architect-review.md
+          format: |
+            ```markdown
+            # Architecture Review
+
+            ## Result: APPROVE / IMPROVE / REJECT
+
+            ## Summary
+            {1-2 sentences summarizing result}
+
+            ## Reviewed Perspectives
+            - [x] Structure & Design
+            - [x] Code Quality
+            - [x] Change Scope
+
+            ## Issues (if REJECT)
+            | # | Location | Issue | Fix |
+            |---|----------|-------|-----|
+            | 1 | `src/file.ts:42` | Issue description | Fix method |
+
+            ## Improvement Suggestions (optional, non-blocking)
+            - {Future improvement suggestions}
+            ```
+
+            **Cognitive load reduction rules:**
+            - APPROVE + no issues -> Summary only (5 lines or less)
+            - APPROVE + minor suggestions -> Summary + suggestions (15 lines or less)
+            - REJECT -> Issues in table format (30 lines or less)
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          Focus on **architecture and design** review. Do NOT review AI-specific issues (that's the ai_review step).
+
+          Review the code and provide feedback.
+
+      - name: security-review
+        edit: false
+        agent: ../agents/default/security-reviewer.md
+        report:
+          name: 02-security-review.md
+          format: |
+            ```markdown
+            # Security Review
+
+            ## Result: APPROVE / REJECT
+
+            ## Severity: None / Low / Medium / High / Critical
+
+            ## Check Results
+            | Category | Result | Notes |
+            |----------|--------|-------|
+            | Injection | - | - |
+            | Auth/Authz | - | - |
+            | Data Protection | - | - |
+            | Dependencies | - | - |
+
+            ## Vulnerabilities (if REJECT)
+            | # | Severity | Type | Location | Fix |
+            |---|----------|------|----------|-----|
+            | 1 | High | SQLi | `src/db.ts:42` | Use parameterized query |
+
+            ## Warnings (non-blocking)
+            - {Security recommendations}
+            ```
+
+            **Cognitive load reduction rules:**
+            - No issues -> Check table only (10 lines or less)
+            - Warnings -> + Warnings 1-2 lines (15 lines or less)
+            - Vulnerabilities -> + Table format (30 lines or less)
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          Perform security review on the code. Check for vulnerabilities including:
+          - Injection attacks (SQL, Command, XSS)
+          - Authentication/Authorization issues
+          - Data exposure risks
+          - Cryptographic weaknesses
+
+      - name: ai-review
+        edit: false
+        agent: ../agents/default/ai-antipattern-reviewer.md
+        report:
+          name: 03-ai-review.md
+          format: |
+            ```markdown
+            # AI-Generated Code Review
+
+            ## Result: APPROVE / REJECT
+
+            ## Summary
+            {One sentence summarizing result}
+
+            ## Verified Items
+            | Aspect | Result | Notes |
+            |--------|--------|-------|
+            | Assumption validity | - | - |
+            | API/Library existence | - | - |
+            | Context fit | - | - |
+            | Scope | - | - |
+
+            ## Issues (if REJECT)
+            | # | Category | Location | Issue |
+            |---|----------|----------|-------|
+            | 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
+            ```
+
+            **Cognitive load reduction rules:**
+            - No issues -> Summary 1 line + check table only (10 lines or less)
+            - Issues found -> + Issues in table format (25 lines or less)
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          Review the code for AI-specific issues:
+          - Assumption validation
+          - Plausible but wrong patterns
+          - Context fit with existing codebase
+          - Scope creep detection
+    rules:
+      - condition: all("approved")
+        next: supervise
+      - condition: any("needs_fix")
+        next: supervise
+
+  - name: supervise
+    edit: false
+    agent: ../agents/default/supervisor.md
+    report:
+      - Review Summary: 04-review-summary.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - Write
+      - WebSearch
+      - WebFetch
+    rules:
+      - condition: approved, PR comment requested
+        next: pr-comment
+      - condition: approved
+        next: COMPLETE
+      - condition: rejected
+        next: ABORT
+    pass_previous_response: true
+    instruction_template: |
+      ## Review Results
+      {previous_response}
+
+      **This is a review-only workflow.** Do NOT run tests or builds.
+      Your role is to synthesize the review results and produce a final summary.
+
+      **Tasks:**
+      1. Read all review reports in the Report Directory
+      2. Synthesize findings from architecture, security, and AI reviews
+      3. Produce a consolidated review summary with overall verdict
+      4. Determine routing:
+         - If the task mentions posting to a PR (e.g., "post comments to PR", "comment on PR"),
+           route to `pr-comment` step (condition: "approved, PR comment requested")
+         - If local review only, route to COMPLETE (condition: "approved")
+         - If critical issues found, route to ABORT (condition: "rejected")
+
+      **Review Summary report format:**
+      ```markdown
+      # Review Summary
+
+      ## Overall Verdict: APPROVE / REJECT
+
+      ## Summary
+      {2-3 sentences consolidating all review results}
+
+      ## Review Results
+      | Review | Result | Key Findings |
+      |--------|--------|--------------|
+      | Architecture | APPROVE/REJECT | {Brief finding} |
+      | Security | APPROVE/REJECT | {Brief finding} |
+      | AI Antipattern | APPROVE/REJECT | {Brief finding} |
+
+      ## Issues Requiring Attention
+      | # | Severity | Source | Location | Issue |
+      |---|----------|--------|----------|-------|
+      | 1 | High | Security | `file:line` | Description |
+
+      ## Improvement Suggestions
+      - {Consolidated suggestions from all reviews}
+      ```
+
+  - name: pr-comment
+    edit: false
+    agent: ../agents/review/pr-commenter.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - Bash
+    rules:
+      - condition: Comments posted
+        next: COMPLETE
+      - condition: Failed to post comments
+        next: COMPLETE
+    pass_previous_response: true
+    instruction_template: |
+      ## Review Summary
+      {previous_response}
+
+      Post the review results to the PR as comments.
+
+      **Procedure:**
+      1. Extract the PR number from the task description
+      2. Read all review reports in the Report Directory:
+         - `01-architect-review.md` (Architecture review)
+         - `02-security-review.md` (Security review)
+         - `03-ai-review.md` (AI antipattern review)
+         - `04-review-summary.md` (Consolidated summary)
+      3. Filter findings by severity and post inline comments for Critical/High/Medium
+      4. Post a summary comment with the following format:
+
+      ```
+      ## Automated Review Summary
+
+      {Overall verdict and summary from 04-review-summary.md}
+
+      ### Review Results
+      | Review | Result |
+      |--------|--------|
+      | Architecture | {result} |
+      | Security | {result} |
+      | AI Antipattern | {result} |
+
+      ### Key Findings
+      {Bulleted list of important findings}
+
+      ### Improvement Suggestions
+      {Consolidated suggestions}
+
+      ---
+      *Generated by [takt](https://github.com/toruticas/takt) review-only workflow*
+      ```
diff --git a/resources/global/ja/agents/review/pr-commenter.md b/resources/global/ja/agents/review/pr-commenter.md
new file mode 100644
index 0000000..be339f2
--- /dev/null
+++ b/resources/global/ja/agents/review/pr-commenter.md
@@ -0,0 +1,75 @@
+# PR Commenter Agent
+
+あなたは**PRコメント投稿の専門家**です。`gh` CLIを使用してレビューの指摘をGitHub Pull Requestに投稿します。
+
+## 役割
+
+- レビューの指摘をPRコメントとして投稿
+- 開発者向けに指摘を明確かつ簡潔にフォーマット
+- 重要度によるフィルタリングでノイズを削減
+
+**やらないこと:**
+- 自分でコードをレビューする（レビュアーが既に実施済み）
+- ファイルの編集
+- テストやビルドの実行
+- コード品質の判断（レビュアーの結果をそのまま投稿する）
+
+## コア知識
+
+### GitHub PR Comment API
+
+**インラインレビューコメント**（ファイル・行番号ごとの指摘）:
+
+```bash
+gh api repos/{owner}/{repo}/pulls/{pr_number}/comments \
+  -f body="**[{category}]** {description}" \
+  -f path="{file_path}" \
+  -F line={line_number} \
+  -f commit_id="$(gh pr view {pr_number} --json headRefOid -q .headRefOid)"
+```
+
+- `commit_id` にはPRのHEADコミットを使用
+- 同一行に複数の指摘がある場合は1つのコメントにまとめる
+
+**サマリーコメント**（全体レビュー）:
+
+```bash
+gh pr comment {pr_number} --body "{markdown_body}"
+```
+
+- 複数行のbodyにはHEREDOCを使用してエスケープ問題を回避
+
+### PR番号の抽出
+
+タスクコンテキストから一般的なパターンでPR番号を抽出:
+- "PR #42"、"#42"、"pull/42"、"pulls/42"
+- PR番号が見つからない場合は報告して投稿せずに終了
+
+## コメント品質の原則
+
+### 重要度ベースのフィルタリング
+
+| 重大度 | アクション |
+|--------|-----------|
+| Critical / High | 必ずインラインコメントとして投稿 |
+| Medium | インラインコメントとして投稿 |
+| Low | サマリーにのみ含める |
+| Informational | サマリーにのみ含める |
+
+### フォーマット
+
+- **簡潔に。** PRコメントはアクション可能で要点を押さえたものにする
+- **場所を含める。** 可能な限り具体的なファイルと行番号を参照
+- **指摘を分類する。** `[Security]`、`[Architecture]`、`[AI Pattern]` のようなラベルを使用
+
+## エラーハンドリング
+
+- `gh` コマンドが失敗した場合はエラーを報告するが、過度にリトライしない
+- PR番号が特定できない場合は情報メッセージを出力して完了
+- 投稿する指摘がない場合はサマリーコメントのみ投稿
+
+## 重要
+
+- **ファイルを変更しない。** コメント投稿のみ行う。
+- **レート制限を尊重。** 個別コメントを大量投稿しない。可能な限りまとめる。
+- **レビューレポートを情報源とする。** 自分の分析ではなく、レビュアーの結果を投稿する。
diff --git a/resources/global/ja/workflows/review-only.yaml b/resources/global/ja/workflows/review-only.yaml
new file mode 100644
index 0000000..14e78b3
--- /dev/null
+++ b/resources/global/ja/workflows/review-only.yaml
@@ -0,0 +1,332 @@
+# レビュー専用ワークフロー
+# コードやPRをレビューするだけで編集は行わない
+# ローカル: コンソール出力のみ。PR指定時: PRにインラインコメント＋サマリを投稿
+#
+# フロー:
+# plan -> reviewers (parallel: arch-review + security-review + ai-review) -> supervise
+#   -> pr-comment -> COMPLETE (PRコメント要求時)
+#   -> COMPLETE (ローカル: コンソール出力のみ)
+#   -> ABORT (rejected)
+#
+# 全ステップ edit: false（ファイル変更なし）
+#
+# テンプレート変数:
+#   {iteration}         - ワークフロー全体のターン数
+#   {max_iterations}    - 最大イテレーション数
+#   {step_iteration}    - ステップごとのイテレーション数
+#   {task}              - 元のユーザー要求
+#   {previous_response} - 前のステップの出力
+#   {user_inputs}       - 蓄積されたユーザー入力
+#   {report_dir}        - レポートディレクトリ名
+
+name: review-only
+description: レビュー専用ワークフロー - コードをレビューするだけで編集は行わない
+
+max_iterations: 10
+
+initial_step: plan
+
+steps:
+  - name: plan
+    edit: false
+    agent: ../agents/default/planner.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - WebSearch
+      - WebFetch
+    rules:
+      - condition: レビュー対象が明確
+        next: reviewers
+      - condition: ユーザーが質問をしている（レビュータスクではない）
+        next: COMPLETE
+      - condition: 要件が不明確、情報不足
+        next: ABORT
+        appendix: |
+          確認事項:
+          - {質問1}
+          - {質問2}
+    pass_previous_response: true
+    instruction_template: |
+      ## Previous Response (superviseからの差し戻し時)
+      {previous_response}
+
+      レビュー依頼を分析し、レビュー方針を立ててください。
+
+      **これはレビュー専用ワークフローです。** コード編集は行いません。
+      以下に集中してください:
+      1. レビュー対象のファイル/モジュールを特定
+      2. レビューの重点領域を決定（アーキテクチャ、セキュリティ、AIパターン等）
+      3. 依頼に記載された特定の懸念事項を整理
+
+      **PR番号が記載されている場合**（例: "PR #42"）、レビュアーが
+      PRの変更ファイルに集中できるよう計画に含めてください。
+
+  - name: reviewers
+    parallel:
+      - name: arch-review
+        edit: false
+        agent: ../agents/default/architecture-reviewer.md
+        report:
+          name: 01-architect-review.md
+          format: |
+            ```markdown
+            # アーキテクチャレビュー
+
+            ## 結果: APPROVE / IMPROVE / REJECT
+
+            ## サマリー
+            {1-2文で結果を要約}
+
+            ## 確認した観点
+            - [x] 構造・設計
+            - [x] コード品質
+            - [x] 変更スコープ
+
+            ## 問題点（REJECTの場合）
+            | # | 場所 | 問題 | 修正案 |
+            |---|------|------|--------|
+            | 1 | `src/file.ts:42` | 問題の説明 | 修正方法 |
+
+            ## 改善提案（任意・ブロッキングではない）
+            - {将来的な改善提案}
+            ```
+
+            **認知負荷軽減ルール:**
+            - APPROVE + 問題なし → サマリーのみ（5行以内）
+            - APPROVE + 軽微な提案 → サマリー + 改善提案（15行以内）
+            - REJECT → 問題点を表形式で（30行以内）
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          **アーキテクチャと設計**のレビューに集中してください。AI特有の問題はレビューしないでください（ai_reviewステップで行います）。
+
+          コードをレビューしてフィードバックを提供してください。
+
+      - name: security-review
+        edit: false
+        agent: ../agents/default/security-reviewer.md
+        report:
+          name: 02-security-review.md
+          format: |
+            ```markdown
+            # セキュリティレビュー
+
+            ## 結果: APPROVE / REJECT
+
+            ## 重大度: None / Low / Medium / High / Critical
+
+            ## チェック結果
+            | カテゴリ | 結果 | 備考 |
+            |---------|------|------|
+            | インジェクション | - | - |
+            | 認証・認可 | - | - |
+            | データ保護 | - | - |
+            | 依存関係 | - | - |
+
+            ## 脆弱性（REJECTの場合）
+            | # | 重大度 | 種類 | 場所 | 修正案 |
+            |---|--------|------|------|--------|
+            | 1 | High | SQLi | `src/db.ts:42` | パラメータ化クエリを使用 |
+
+            ## 警告（ブロッキングではない）
+            - {セキュリティに関する推奨事項}
+            ```
+
+            **認知負荷軽減ルール:**
+            - 問題なし → チェック表のみ（10行以内）
+            - 警告あり → + 警告を1-2行（15行以内）
+            - 脆弱性あり → + 表形式で（30行以内）
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          コードに対してセキュリティレビューを行ってください。以下の脆弱性を確認してください:
+          - インジェクション攻撃（SQL, コマンド, XSS）
+          - 認証・認可の問題
+          - データ露出リスク
+          - 暗号化の弱点
+
+      - name: ai-review
+        edit: false
+        agent: ../agents/default/ai-antipattern-reviewer.md
+        report:
+          name: 03-ai-review.md
+          format: |
+            ```markdown
+            # AI生成コードレビュー
+
+            ## 結果: APPROVE / REJECT
+
+            ## サマリー
+            {1文で結果を要約}
+
+            ## 検証した項目
+            | 観点 | 結果 | 備考 |
+            |------|------|------|
+            | 仮定の妥当性 | - | - |
+            | API/ライブラリの実在 | - | - |
+            | コンテキスト適合 | - | - |
+            | スコープ | - | - |
+
+            ## 問題点（REJECTの場合）
+            | # | カテゴリ | 場所 | 問題 |
+            |---|---------|------|------|
+            | 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
+            ```
+
+            **認知負荷軽減ルール:**
+            - 問題なし → サマリー1文 + チェック表のみ（10行以内）
+            - 問題あり → + 問題を表形式で（25行以内）
+        allowed_tools:
+          - Read
+          - Glob
+          - Grep
+          - Write
+          - WebSearch
+          - WebFetch
+        rules:
+          - condition: approved
+          - condition: needs_fix
+        instruction_template: |
+          AI特有の問題についてコードをレビューしてください:
+          - 仮定の検証
+          - もっともらしいが間違っているパターン
+          - 既存コードベースとの適合性
+          - スコープクリープの検出
+    rules:
+      - condition: all("approved")
+        next: supervise
+      - condition: any("needs_fix")
+        next: supervise
+
+  - name: supervise
+    edit: false
+    agent: ../agents/default/supervisor.md
+    report:
+      - Review Summary: 04-review-summary.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - Write
+      - WebSearch
+      - WebFetch
+    rules:
+      - condition: approved, PR comment requested
+        next: pr-comment
+      - condition: approved
+        next: COMPLETE
+      - condition: rejected
+        next: ABORT
+    pass_previous_response: true
+    instruction_template: |
+      ## レビュー結果
+      {previous_response}
+
+      **これはレビュー専用ワークフローです。** テスト実行やビルドは行わないでください。
+      レビュー結果を統合し、最終サマリーを作成する役割です。
+
+      **やること:**
+      1. Report Directory内の全レビューレポートを読む
+      2. アーキテクチャ・セキュリティ・AIレビューの結果を統合
+      3. 統合レビューサマリーと総合判定を作成
+      4. ルーティング判断:
+         - タスクにPRへのコメント投稿が含まれている場合
+           （例: "PRにコメントして"、"PRにレビュー結果を投稿"）
+           → `pr-comment` ステップへ（condition: "approved, PR comment requested"）
+         - ローカルレビューのみ → COMPLETE（condition: "approved"）
+         - 重大な問題が見つかった場合 → ABORT（condition: "rejected"）
+
+      **Review Summaryレポートフォーマット:**
+      ```markdown
+      # レビューサマリー
+
+      ## 総合判定: APPROVE / REJECT
+
+      ## サマリー
+      {2-3文で全レビュー結果を統合}
+
+      ## レビュー結果
+      | レビュー | 結果 | 主要な発見 |
+      |---------|------|-----------|
+      | アーキテクチャ | APPROVE/REJECT | {概要} |
+      | セキュリティ | APPROVE/REJECT | {概要} |
+      | AIアンチパターン | APPROVE/REJECT | {概要} |
+
+      ## 要注意の問題
+      | # | 重大度 | ソース | 場所 | 問題 |
+      |---|--------|--------|------|------|
+      | 1 | High | セキュリティ | `file:line` | 説明 |
+
+      ## 改善提案
+      - {全レビューからの統合提案}
+      ```
+
+  - name: pr-comment
+    edit: false
+    agent: ../agents/review/pr-commenter.md
+    allowed_tools:
+      - Read
+      - Glob
+      - Grep
+      - Bash
+    rules:
+      - condition: コメント投稿完了
+        next: COMPLETE
+      - condition: コメント投稿失敗
+        next: COMPLETE
+    pass_previous_response: true
+    instruction_template: |
+      ## レビューサマリー
+      {previous_response}
+
+      レビュー結果をPRにコメントとして投稿してください。
+
+      **手順:**
+      1. タスク説明からPR番号を抽出
+      2. Report Directory内の全レビューレポートを読む:
+         - `01-architect-review.md`（アーキテクチャレビュー）
+         - `02-security-review.md`（セキュリティレビュー）
+         - `03-ai-review.md`（AIアンチパターンレビュー）
+         - `04-review-summary.md`（統合サマリー）
+      3. 重要度でフィルタリングし、Critical/High/Mediumの指摘をインラインコメントとして投稿
+      4. 以下のフォーマットでサマリーコメントを投稿:
+
+      ```
+      ## 自動レビューサマリー
+
+      {04-review-summary.mdからの総合判定とサマリー}
+
+      ### レビュー結果
+      | レビュー | 結果 |
+      |---------|------|
+      | アーキテクチャ | {結果} |
+      | セキュリティ | {結果} |
+      | AIアンチパターン | {結果} |
+
+      ### 主要な発見
+      {重要な指摘のリスト}
+
+      ### 改善提案
+      {統合された提案}
+
+      ---
+      *[takt](https://github.com/toruticas/takt) review-only ワークフローで生成*
+      ```
diff --git a/src/__tests__/review-only-workflow.test.ts b/src/__tests__/review-only-workflow.test.ts
new file mode 100644
index 0000000..617c790
--- /dev/null
+++ b/src/__tests__/review-only-workflow.test.ts
@@ -0,0 +1,250 @@
+/**
+ * Tests for review-only workflow
+ *
+ * Covers:
+ * - Workflow YAML files (EN/JA) load and pass schema validation
+ * - Workflow structure: plan -> reviewers (parallel) -> supervise -> pr-comment
+ * - All steps have edit: false
+ * - pr-commenter agent has Bash in allowed_tools
+ * - Routing rules for local vs PR comment flows
+ */
+
+import { describe, it, expect } from 'vitest';
+import { readFileSync } from 'node:fs';
+import { join } from 'node:path';
+import { parse as parseYaml } from 'yaml';
+import { WorkflowConfigRawSchema } from '../models/schemas.js';
+
+const RESOURCES_DIR = join(import.meta.dirname, '../../resources/global');
+
+function loadReviewOnlyYaml(lang: 'en' | 'ja') {
+  const filePath = join(RESOURCES_DIR, lang, 'workflows', 'review-only.yaml');
+  const content = readFileSync(filePath, 'utf-8');
+  return parseYaml(content);
+}
+
+describe('review-only workflow (EN)', () => {
+  const raw = loadReviewOnlyYaml('en');
+
+  it('should pass schema validation', () => {
+    const result = WorkflowConfigRawSchema.safeParse(raw);
+    expect(result.success).toBe(true);
+  });
+
+  it('should have correct name and initial_step', () => {
+    expect(raw.name).toBe('review-only');
+    expect(raw.initial_step).toBe('plan');
+  });
+
+  it('should have max_iterations of 10', () => {
+    expect(raw.max_iterations).toBe(10);
+  });
+
+  it('should have 4 steps: plan, reviewers, supervise, pr-comment', () => {
+    const stepNames = raw.steps.map((s: { name: string }) => s.name);
+    expect(stepNames).toEqual(['plan', 'reviewers', 'supervise', 'pr-comment']);
+  });
+
+  it('should have all steps with edit: false', () => {
+    for (const step of raw.steps) {
+      if (step.edit !== undefined) {
+        expect(step.edit).toBe(false);
+      }
+      if (step.parallel) {
+        for (const sub of step.parallel) {
+          if (sub.edit !== undefined) {
+            expect(sub.edit).toBe(false);
+          }
+        }
+      }
+    }
+  });
+
+  it('should have reviewers step with 3 parallel sub-steps', () => {
+    const reviewers = raw.steps.find((s: { name: string }) => s.name === 'reviewers');
+    expect(reviewers).toBeDefined();
+    expect(reviewers.parallel).toHaveLength(3);
+
+    const subNames = reviewers.parallel.map((s: { name: string }) => s.name);
+    expect(subNames).toEqual(['arch-review', 'security-review', 'ai-review']);
+  });
+
+  it('should have reviewers step with aggregate rules', () => {
+    const reviewers = raw.steps.find((s: { name: string }) => s.name === 'reviewers');
+    expect(reviewers.rules).toHaveLength(2);
+    expect(reviewers.rules[0].condition).toBe('all("approved")');
+    expect(reviewers.rules[0].next).toBe('supervise');
+    expect(reviewers.rules[1].condition).toBe('any("needs_fix")');
+    expect(reviewers.rules[1].next).toBe('supervise');
+  });
+
+  it('should have supervise step with routing rules for local and PR flows', () => {
+    const supervise = raw.steps.find((s: { name: string }) => s.name === 'supervise');
+    expect(supervise.rules).toHaveLength(3);
+
+    const conditions = supervise.rules.map((r: { condition: string }) => r.condition);
+    expect(conditions).toContain('approved, PR comment requested');
+    expect(conditions).toContain('approved');
+    expect(conditions).toContain('rejected');
+
+    const prRule = supervise.rules.find((r: { condition: string }) => r.condition === 'approved, PR comment requested');
+    expect(prRule.next).toBe('pr-comment');
+
+    const localRule = supervise.rules.find((r: { condition: string }) => r.condition === 'approved');
+    expect(localRule.next).toBe('COMPLETE');
+
+    const rejectRule = supervise.rules.find((r: { condition: string }) => r.condition === 'rejected');
+    expect(rejectRule.next).toBe('ABORT');
+  });
+
+  it('should have pr-comment step with Bash in allowed_tools', () => {
+    const prComment = raw.steps.find((s: { name: string }) => s.name === 'pr-comment');
+    expect(prComment).toBeDefined();
+    expect(prComment.allowed_tools).toContain('Bash');
+  });
+
+  it('should have pr-comment step using pr-commenter agent', () => {
+    const prComment = raw.steps.find((s: { name: string }) => s.name === 'pr-comment');
+    expect(prComment.agent).toContain('review/pr-commenter.md');
+  });
+
+  it('should have plan step reusing default planner agent', () => {
+    const plan = raw.steps.find((s: { name: string }) => s.name === 'plan');
+    expect(plan.agent).toContain('default/planner.md');
+  });
+
+  it('should have supervise step reusing default supervisor agent', () => {
+    const supervise = raw.steps.find((s: { name: string }) => s.name === 'supervise');
+    expect(supervise.agent).toContain('default/supervisor.md');
+  });
+
+  it('should not have any step with edit: true', () => {
+    for (const step of raw.steps) {
+      expect(step.edit).not.toBe(true);
+      if (step.parallel) {
+        for (const sub of step.parallel) {
+          expect(sub.edit).not.toBe(true);
+        }
+      }
+    }
+  });
+
+  it('reviewer sub-steps should not have Bash in allowed_tools', () => {
+    const reviewers = raw.steps.find((s: { name: string }) => s.name === 'reviewers');
+    for (const sub of reviewers.parallel) {
+      expect(sub.allowed_tools).not.toContain('Bash');
+    }
+  });
+});
+
+describe('review-only workflow (JA)', () => {
+  const raw = loadReviewOnlyYaml('ja');
+
+  it('should pass schema validation', () => {
+    const result = WorkflowConfigRawSchema.safeParse(raw);
+    expect(result.success).toBe(true);
+  });
+
+  it('should have correct name and initial_step', () => {
+    expect(raw.name).toBe('review-only');
+    expect(raw.initial_step).toBe('plan');
+  });
+
+  it('should have same step structure as EN version', () => {
+    const stepNames = raw.steps.map((s: { name: string }) => s.name);
+    expect(stepNames).toEqual(['plan', 'reviewers', 'supervise', 'pr-comment']);
+  });
+
+  it('should have reviewers step with 3 parallel sub-steps', () => {
+    const reviewers = raw.steps.find((s: { name: string }) => s.name === 'reviewers');
+    expect(reviewers.parallel).toHaveLength(3);
+
+    const subNames = reviewers.parallel.map((s: { name: string }) => s.name);
+    expect(subNames).toEqual(['arch-review', 'security-review', 'ai-review']);
+  });
+
+  it('should have all steps with edit: false or undefined', () => {
+    for (const step of raw.steps) {
+      expect(step.edit).not.toBe(true);
+      if (step.parallel) {
+        for (const sub of step.parallel) {
+          expect(sub.edit).not.toBe(true);
+        }
+      }
+    }
+  });
+
+  it('should have pr-comment step with Bash in allowed_tools', () => {
+    const prComment = raw.steps.find((s: { name: string }) => s.name === 'pr-comment');
+    expect(prComment.allowed_tools).toContain('Bash');
+  });
+
+  it('should have same aggregate rules on reviewers', () => {
+    const reviewers = raw.steps.find((s: { name: string }) => s.name === 'reviewers');
+    expect(reviewers.rules[0].condition).toBe('all("approved")');
+    expect(reviewers.rules[1].condition).toBe('any("needs_fix")');
+  });
+});
+
+describe('pr-commenter agent files', () => {
+  it('should exist for EN with domain knowledge', () => {
+    const filePath = join(RESOURCES_DIR, 'en', 'agents', 'review', 'pr-commenter.md');
+    const content = readFileSync(filePath, 'utf-8');
+    expect(content).toContain('PR Commenter');
+    expect(content).toContain('gh api');
+    expect(content).toContain('gh pr comment');
+  });
+
+  it('should exist for JA with domain knowledge', () => {
+    const filePath = join(RESOURCES_DIR, 'ja', 'agents', 'review', 'pr-commenter.md');
+    const content = readFileSync(filePath, 'utf-8');
+    expect(content).toContain('PR Commenter');
+    expect(content).toContain('gh api');
+    expect(content).toContain('gh pr comment');
+  });
+
+  it('should NOT contain workflow-specific report names (EN)', () => {
+    const filePath = join(RESOURCES_DIR, 'en', 'agents', 'review', 'pr-commenter.md');
+    const content = readFileSync(filePath, 'utf-8');
+    // Agent should not reference specific review-only workflow report files
+    expect(content).not.toContain('01-architect-review.md');
+    expect(content).not.toContain('02-security-review.md');
+    expect(content).not.toContain('03-ai-review.md');
+    expect(content).not.toContain('04-review-summary.md');
+    // Agent should not reference specific reviewer names from review-only workflow
+    expect(content).not.toContain('Architecture review report');
+    expect(content).not.toContain('Security review report');
+    expect(content).not.toContain('AI antipattern review report');
+  });
+
+  it('should NOT contain workflow-specific report names (JA)', () => {
+    const filePath = join(RESOURCES_DIR, 'ja', 'agents', 'review', 'pr-commenter.md');
+    const content = readFileSync(filePath, 'utf-8');
+    expect(content).not.toContain('01-architect-review.md');
+    expect(content).not.toContain('02-security-review.md');
+    expect(content).not.toContain('03-ai-review.md');
+    expect(content).not.toContain('04-review-summary.md');
+  });
+});
+
+describe('pr-comment instruction_template contains workflow-specific procedures', () => {
+  it('EN: should reference specific report files', () => {
+    const raw = loadReviewOnlyYaml('en');
+    const prComment = raw.steps.find((s: { name: string }) => s.name === 'pr-comment');
+    const template = prComment.instruction_template;
+    expect(template).toContain('01-architect-review.md');
+    expect(template).toContain('02-security-review.md');
+    expect(template).toContain('03-ai-review.md');
+    expect(template).toContain('04-review-summary.md');
+  });
+
+  it('JA: should reference specific report files', () => {
+    const raw = loadReviewOnlyYaml('ja');
+    const prComment = raw.steps.find((s: { name: string }) => s.name === 'pr-comment');
+    const template = prComment.instruction_template;
+    expect(template).toContain('01-architect-review.md');
+    expect(template).toContain('02-security-review.md');
+    expect(template).toContain('03-ai-review.md');
+    expect(template).toContain('04-review-summary.md');
+  });
+});