Repository navigation
refactor: reorganize skills visibility and naming - #36
Conversation
Migrated all custom commands to skills following Claude Code's official Skills documentation recommendation that "custom slash commands have been merged into skills." Migration Summary: - 35 commands migrated to skills - Created 12 new skills with references: - work, planning, harness-init, ultrawork (core) - sync-status, cc-cursor-cc (core utilities) - setup-tools (consolidates all setup commands) - harness-update, harness-mem, harness-ui - handoff (for 2-agent workflow) - Enhanced existing skills: - codex-review (added codex-mode) - harness-review (added hooks) - session (added subcommands) - Updated rules: - Added skill-editing.md - Deprecated command-editing.md Benefits: - Skills have more features (auto-loading, supporting files) - Unified architecture following official docs - Better organization with references/ subdirectories - Cleaner separation of concerns https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
Self-review found 5 commands that were deleted but not migrated: - crud.md → New skills/crud/ skill - localize-rules.md → New skills/localize-rules/ skill - release.md → New skills/release/ skill - sync-project-specs.md → Added to skills/memory/references/ - sync-ssot-from-memory.md → Added to skills/memory/references/ Also updated skills/memory/SKILL.md to reference the sync commands. https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
Per official Claude Code Skills documentation, Task content skills (those with side effects like file generation, commits, deployments) should have `disable-model-invocation: true` to prevent Claude from auto-invoking them based on description matching. Added to: - work (executes Plans.md, commits) - planning (creates Plans.md) - harness-init (generates project structure) - crud (generates CRUD code) - setup-tools (installs packages, modifies config) - localize-rules (modifies rule files) - release (creates tags, pushes) - deploy (deploys to production) Reference: https://code.claude.com/docs/en/skills#types-of-skill-content https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
- Hide internal skills (user-invocable: false): crud, auth, impl, plans-management, verify, harness-ui, ui - Delete unused skill: cc-cursor-cc - Show user-facing skills: handoff, session-memory - Rename skills for clarity: dev-browser → agent-browser docs → notebookLM video → generate-video - Merge workflow into handoff (consolidate handoff functionality) https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
Internal skill auto-loaded by description triggers, not directly invoked. https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
crud had both disable-model-invocation: true and user-invocable: false, making it completely inaccessible. Removed disable-model-invocation so Claude can auto-invoke when user mentions CRUD generation. https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
- Add user-invocable: false to session-control, session-state (internal) - Remove non-standard version/dependencies from session-state - Add "Do NOT load for" to 11 skills to prevent false positives: - agent-browser: sharing URLs, embedding links, screenshot files - crud: UI creation, form design, DB schema discussions - generate-video: video players, live demos, playback features - handoff: casual completion, progress chat, informal updates - harness-mem: app memory features, casual remember requests - harness-update: app updates, dependency updates - localize-rules: app i18n, business rules - release: planning discussions, version mentions, casual talk - session: app user sessions, login state - setup-tools: app setup screens, onboarding, CI questions - sync-status: casual progress chat https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
- Restore version/dependencies to session-state (used by validate-skills.sh) - Add "Do NOT load for" to internal skills: - session-state: user session management, login state, app state - session-control: user session management, login state, app state - harness-ui: app UI implementation, dashboard components, admin panels https://claude.ai/code/session_01BRUvdz9BZmbRBeXierHWRR
|
Caution Review failedThe pull request is closed. WalkthroughこのPRは、複数のスキルをリネームし、新しいスキル文書ファイルを作成し、古いコマンド文書ファイルを削除することで、プロジェクト構造を再編成しています。これはコマンドベースからスキルベースのアーキテクチャへの移行の一部です。 Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
Note
Due to the large number of review comments, Critical, Major severity comments were prioritized as inline comments.
🤖 Fix all issues with AI agents
In `@skills/memory/references/sync-project-specs.md`:
- Around line 87-88: The doc mentions a non-existent `/remember` command; either
implement the `/remember` command (resolve or link to tracking item `#10350` and
add its command handler and docs) or remove/replace its mention in the project
spec text: edit the "project-specific operations" guidance in the
sync-project-specs content to delete the `/remember` line or swap it for an
existing command such as `/sync-status` (and update the surrounding wording
accordingly) so the document no longer instructs users to run an unimplemented
`/remember`.
In `@skills/setup-tools/references/webhook-setup.md`:
- Around line 67-71: 現在のhere-doc uses single-quoted delimiter << 'EOF' which
prevents shell variable expansion so CHANGED_FILES is passed literally; change
the heredoc to an unquoted delimiter (<<EOF) so $CHANGED_FILES expands, and
ensure the command inside uses proper quoting like /harness-review --ci --files
"$CHANGED_FILES" (refer to the CHANGED_FILES assignment and the heredoc block
using << 'EOF' in the snippet) to pass the actual file list to the
harness-review command.
🟡 Minor comments (23)
skills/codex-review/references/codex-mode.md-34-37 (1)
34-37:⚠️ Potential issue | 🟡 Minor「9 Experts」表記と列挙数が不一致です。
リストは8件なので、見出しの数と一致させてください（例: 見出しを8に修正 or 1名追加）。
Also applies to: 49-65
skills/setup-tools/references/remotion-setup.md-152-161 (1)
152-161:⚠️ Potential issue | 🟡 Minorナレーション生成スクリプトのパスが手順と不一致
Step 6 で
remotion/src/utils/...を作るのに、generate-narrationはsrc/utils/...を指していて実行失敗しやすいです。remotion/直下に合わせてください。🧩 修正案
"scripts": { "remotion": "remotion studio remotion/index.ts", "render": "remotion render remotion/index.ts Main out/video.mp4", "render:gif": "remotion render remotion/index.ts Main out/video.gif", - "generate-narration": "npx ts-node src/utils/narration-generator.ts" + "generate-narration": "npx ts-node remotion/src/utils/narration-generator.ts" }skills/setup-tools/references/remotion-setup.md-110-114 (1)
110-114:⚠️ Potential issue | 🟡 Minorフェンスコードに言語指定が不足
lint（MD040）に引っかかるので
bash指定を追加してください。✍️ 修正案
-``` +```bash PLUGIN_DIR="${CLAUDE_PLUGIN_ROOT:-$(dirname $(dirname $0))}" cp -r "$PLUGIN_DIR/templates/remotion/"* remotion/</details> </blockquote></details> <details> <summary>skills/setup-tools/SKILL.md-32-41 (1)</summary><blockquote> `32-41`: _⚠️ Potential issue_ | _🟡 Minor_ **サブコマンド表に`ui`と`skills`が欠落しています。** `argument-hint`に含まれるため、一覧にも追記して整合性を取ってください。 </blockquote></details> <details> <summary>skills/setup-tools/references/opencode-setup.md-78-85 (1)</summary><blockquote> `78-85`: _⚠️ Potential issue_ | _🟡 Minor_ **スキル名がPRのリネーム内容と不整合です。** `docs`はPR目的で`notebookLM`へ変更されるため、完了メッセージのスキル一覧を最新名称に合わせて更新してください。 </blockquote></details> <details> <summary>skills/setup-tools/references/dev-tools-setup.md-41-43 (1)</summary><blockquote> `41-43`: _⚠️ Potential issue_ | _🟡 Minor_ **Python検出コマンドが片方のファイル欠落で検出漏れになります。** `ls requirements.txt pyproject.toml 2>/dev/null` はどちらか一方が無いと終了コード2（非0）になり、`echo "Python detected"` が実行されません。requirements.txtまたはpyproject.tomlのどちらか一方が存在すれば検出できるよう修正してください。 <details> <summary>修正案</summary> ```diff -ls requirements.txt pyproject.toml 2>/dev/null && echo "Python detected" +([ -f requirements.txt ] || [ -f pyproject.toml ]) && echo "Python detected"skills/release/SKILL.md-141-144 (1)
141-144:⚠️ Potential issue | 🟡 MinorRelated Skills の
docsは新名称に更新してください。PR で
docs → notebookLMに改名されているため、参照名を更新しないと導線が切れます。修正案
-- `docs` - Documentation updates +- `notebookLM` - Documentation updatesskills/harness-init/references/execution-flow.md-48-56 (1)
48-56:⚠️ Potential issue | 🟡 Minorフェンスコードに言語指定を追加してください。
markdownlint の MD040 に該当しています。
textなどの指定で十分です。修正案
-``` +```text Combine Step 1 response + background analysis result ...</details> </blockquote></details> <details> <summary>skills/planning/SKILL.md-35-38 (1)</summary><blockquote> `35-38`: _⚠️ Potential issue_ | _🟡 Minor_ **`/plan-with-cc` の表記を `/plan-with-agent` に統一してください。** 2-agent mode のコマンドが学習内容と不一致で、Plans.md への誘導がブレます。表記を統一して誤誘導を防いでください。 Based on learnings: “Use `/plan-with-agent` to add tasks to Plans.md, `/work` to implement tasks, `/harness-review` for quality checks, and `./tests/validate-plugin.sh` for structure validation”. </blockquote></details> <details> <summary>skills/work/references/review-loop.md-7-40 (1)</summary><blockquote> `7-40`: _⚠️ Potential issue_ | _🟡 Minor_ **フェンスコードの言語指定を追加してください。** markdownlint の MD040 対応が必要です。図示用なら `text` でOKです。 <details> <summary>修正案</summary> ```diff -``` +```text /work --parallel 3 ...```diff -``` +```text ⚠️ Escalation (2 items) ...</details> Also applies to: 46-59 </blockquote></details> <details> <summary>skills/harness-update/references/version-detection.md-29-46 (1)</summary><blockquote> `29-46`: _⚠️ Potential issue_ | _🟡 Minor_ **PLUGIN_ROOT の参照先を Step 2 と Step 2.5 で統一してください。** `CLAUDE_PLUGIN_ROOT/claude-code-harness` と `PLUGIN_ROOT` の定義がズレると、比較対象が変わるリスクがあります。Step 2 でも `PLUGIN_ROOT` を使う形に揃えるのが安全です。 <details> <summary>修正案</summary> ```diff -PLUGIN_VERSION=$(cat "$CLAUDE_PLUGIN_ROOT/claude-code-harness/VERSION" 2>/dev/null || echo "unknown") +PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/plugins/claude-code-harness}" +PLUGIN_VERSION=$(cat "$PLUGIN_ROOT/VERSION" 2>/dev/null || echo "unknown")skills/work/SKILL.md-35-36 (1)
35-36:⚠️ Potential issue | 🟡 Minormarkdownlint の指摘（表の空白・コードブロック言語）を解消してください。
テーブルのパイプ前後スペースとコードフェンスの言語指定が欠けているため、lint を通す場合は修正が必要です。
Also applies to: 50-50, 70-72, 81-82, 90-91, 108-109
skills/work/references/execution-flow.md-7-7 (1)
7-7:⚠️ Potential issue | 🟡 Minormarkdownlint の指摘（表の空白・コードブロック言語）を解消してください。
フェンス言語指定とテーブルの空白が不足しているため、lint 通過のため修正が必要です。
Also applies to: 25-25, 32-32, 50-50, 63-63, 81-81, 107-107, 116-116, 133-133
skills/work/references/session-management.md-43-50 (1)
43-50:⚠️ Potential issue | 🟡 Minorマーカー名称の整合性を明記して混乱を避けてください。
pm:requested/cursor:requestedの記載は、既存の運用ルールで言及されているpm:依頼中と混同されやすいので、同義・変換ルールの明記か表記統一をお願いします。
Based on learnings: "Execute/handoff-to-cursorcommand after/workcompletion and review approval when in 2-Agent mode (when Plans.md containspm:依頼中)".skills/work/references/execution-flow.md-129-139 (1)
129-139:⚠️ Potential issue | 🟡 Minor2-Agent ハンドオフの宛先を明確化してください。
ここでは
/handoff-to-cursorのみ記載されていますが、2-Agent では/handoff-to-opencodeも必要な運用があるため、条件分岐または両方の記載をお願いします。
Based on learnings: "Execute/handoff-to-opencodecommand after/workcompletion and review approval when in 2-Agent mode (when Plans.md containspm:依頼中)".skills/work/references/parallel-execution.md-14-19 (1)
14-19:⚠️ Potential issue | 🟡 Minormarkdownlint の指摘（表揃え・コードブロック言語）を修正してください。
テーブル列の整列とコードフェンスの言語指定が不足しているため、lint を通す場合は修正が必要です。
Also applies to: 47-48, 56-56, 64-64
skills/work/references/auto-commit.md-5-5 (1)
5-5:⚠️ Potential issue | 🟡 Minorコードブロックに言語指定を付与してください。
markdownlint の指摘通り、フェンスに言語が無いので修正お願いします。
Also applies to: 16-16, 37-37, 56-56, 73-73, 90-90
skills/work/references/auto-commit.md-88-99 (1)
88-99:⚠️ Potential issue | 🟡 Minor2-Agent のハンドオフ先を両方明記してください。
2-Agent 時の handoff は
/handoff-to-cursorだけでなく/handoff-to-opencodeも対象になるため、どちらを使うか（条件/優先順位）を明示してください。
Based on learnings: "Execute/handoff-to-opencodecommand after/workcompletion and review approval when in 2-Agent mode (when Plans.md containspm:依頼中)".skills/work/references/error-handling.md-5-5 (1)
5-5:⚠️ Potential issue | 🟡 Minormarkdownlint の指摘（表の空白・コードブロック言語）を修正してください。
テーブルの整列とフェンス言語指定が不足しているため、lint を通す場合は修正が必要です。
Also applies to: 28-28, 47-58, 62-62, 71-71
skills/harness-update/references/backup-and-update.md-192-206 (1)
192-206:⚠️ Potential issue | 🟡 Minor変数名の不整合と未定義変数を修正してください。
CLAUDE_PLUGIN_ROOTとPLUGIN_ROOTが混在し、CURRENT_VERSIONが未定義です。スニペット実行時に誤動作するため統一してください。✅ 修正案
-PLUGIN_VERSION=$(cat "$CLAUDE_PLUGIN_ROOT/claude-code-harness/VERSION") +PLUGIN_VERSION=$(cat "$PLUGIN_ROOT/VERSION") -cat > .claude-code-harness-version <<EOF +cat > .claude-code-harness-version <<EOF # claude-code-harness version tracking # This file is auto-generated by /harness-update # DO NOT manually edit - used for update detection version: $PLUGIN_VERSION installed_at: $(grep "^installed_at:" .claude-code-harness-version | cut -d' ' -f2) updated_at: $(date +%Y-%m-%d) last_setup_command: harness-update EOF -echo "✅ Version updated: v$CURRENT_VERSION → v$PLUGIN_VERSION" +echo "✅ Version updated: v${CURRENT_VERSION:-unknown} → v$PLUGIN_VERSION"skills/harness-update/references/verification.md-50-65 (1)
50-65:⚠️ Potential issue | 🟡 Minorコードフェンスに言語指定を追加してください（lint 警告）。
MD040 対応です。✍️ 修正例
-``` +```text 📊 Update Report ...</details> </blockquote></details> <details> <summary>skills/harness-update/references/verification.md-27-29 (1)</summary><blockquote> `27-29`: _⚠️ Potential issue_ | _🟡 Minor_ **テーブルのパイプ形式と列数を修正してください（lint 警告）。** markdownlint の MD055/MD056 に該当します。 <details> <summary>🧹 修正例</summary> ```diff -> | File | Status | Reason | -> |------|--------|--------| -> {{remaining files list}} +> | File | Status | Reason | +> |------|--------|--------| +> | {{file}} | {{status}} | {{reason}} |-> | File | Result | Method | -> |------|--------|--------| -> {{file-by-file results list}} +> | File | Result | Method | +> |------|--------|--------| +> | {{file}} | {{result}} | {{method}} |Also applies to: 79-81
skills/harness-update/references/verification.md-12-14 (1)
12-14:⚠️ Potential issue | 🟡 Minor
jq前提のままなのでガードを追加してください。
Step 1 のスニペットはjqが無い環境で失敗するため、事前チェックか注記を入れると安全です。💡 例: `jq` チェック追加
-REMAINING_UPDATES=$(echo "$VERIFY_RESULT" | jq -r '.updatesCount // 0') -REMAINING_INSTALLS=$(echo "$VERIFY_RESULT" | jq -r '.installsCount // 0') +if command -v jq >/dev/null 2>&1; then + REMAINING_UPDATES=$(echo "$VERIFY_RESULT" | jq -r '.updatesCount // 0') + REMAINING_INSTALLS=$(echo "$VERIFY_RESULT" | jq -r '.installsCount // 0') +else + echo "⚠️ jq が見つかりません。`jq` をインストールして再実行してください。" + REMAINING_UPDATES=0 + REMAINING_INSTALLS=0 +fi
🧹 Nitpick comments (24)
skills/crud/SKILL.md (3)
33-41: 親スキル/子スキルの階層ルーティングを明示しておくと運用が安定します。このセクションに「このスキルはどの親カテゴリに属し、どの意図でルーティングされるか」を明記すると、階層構造のガイドラインに沿った運用がしやすくなります。
Based on learnings: “Organize skills in hierarchical structure with parent skills (categories) and child skills (specific functionality). Parent skills route to appropriate child skills based on user intent and context”.
37-40: テーブル整列のルール違反を解消してください。markdownlint の MD060 によりテーブル列の整列が指摘されています。空白の調整で解消できます。
🔧 修正案（表の整列）
-| Skill | Purpose | When to Call | -|-------|---------|--------------| -| `impl` | Implementation (parent skill) | CRUD feature implementation | -| `verify` | Verification (parent skill) | Post-implementation verification | +| Skill | Purpose | When to Call | +|--------|----------------------------------|-------------------------------| +| `impl` | Implementation (parent skill) | CRUD feature implementation | +| `verify` | Verification (parent skill) | Post-implementation verification |-| Framework | Detection | Generated Files | -|-----------|-----------|-----------------| -| **Next.js + Prisma** | `prisma/schema.prisma` | API routes, Prisma client | -| **Next.js + Drizzle** | `drizzle.config.ts` | API routes, Drizzle queries | -| **Express** | `express` in package.json | Controllers, routes | -| **Hono** | `hono` in package.json | Route handlers | +| Framework | Detection | Generated Files | +|--------------------|-------------------------|------------------------------| +| **Next.js + Prisma** | `prisma/schema.prisma` | API routes, Prisma client | +| **Next.js + Drizzle** | `drizzle.config.ts` | API routes, Drizzle queries | +| **Express** | `express` in package.json | Controllers, routes | +| **Hono** | `hono` in package.json | Route handlers |Also applies to: 77-82
88-99: コードフェンスに言語指定を追加してください。markdownlint の MD040 で言語指定なしが指摘されています。
🔧 修正案
-``` +```text src/ ├── lib/ │ └── validations/ │ └── {entity}.ts # Zod schemas ├── app/api/{entity}/ │ ├── route.ts # GET (list), POST (create) │ └── [id]/ │ └── route.ts # GET, PUT, DELETE └── tests/ └── {entity}.test.ts # Test cases</details> </blockquote></details> <details> <summary>skills/handoff/SKILL.md (2)</summary><blockquote> `14-20`: **テーブルのパイプ周りのスペースを揃えてください（markdownlint MD060）** 現状の区切り行がコンパクトスタイルの要件を満たしていないため、スペースを追加してください。 <details> <summary>修正案</summary> ```diff -|------|------| +| ------ | ------ |-|-----------|----------|--------------| +| ----------- | ---------- | -------------- |Also applies to: 32-36
38-41: 強調テキストではなく見出し化してください（markdownlint MD036）
セクション見出しとして扱いたい意図なら、見出し記法に変更する方が明確です。修正案
-**Why handoff requires review approval:** +### Why handoff requires review approvalskills/session/SKILL.md (1)
27-34: フェンスドコードに言語指定を追加してください。markdownlint の MD040 に該当しており、言語タグがないコードブロックが 2 か所あります。
textなどを付けると警告が解消されます。✅ 例（text を指定）
-``` +```text 📋 Active Sessions ... -``` +``` -``` +```text 📬 Session Inbox ... -``` +```Also applies to: 40-47
skills/sync-status/SKILL.md (1)
42-42: Markdownフォーマットの軽微な問題。静的解析ツールが以下を検出しています：
- 行42, 89: テーブルのパイプ記号の前後にスペースが不足
- 行51, 77: コードブロックに言語指定子が欠落
📝 フォーマット改善の提案
テーブルのスペース修正:
-| Check Item | Detection Method | -|------------|------------------| +| Check Item | Detection Method | +|---------------------|--------------------------| | Done but still `cc:WIP` | Commit history vs marker |-| Situation | Warning | -|-----------|---------| +| Situation | Warning | +|-------------------------------|-------------------------------------| | Multiple `cc:WIP` | ⚠️ Multiple tasks in progress |言語指定子の追加:
-``` +```markdown 📝 Plans.md update needed-``` +```text 🎯 What to do nextAlso applies to: 51-51, 77-77, 89-89
.claude/rules/command-editing.md (1)
1-6: 非推奨化の表示が明確で適切です。レガシードキュメントを参照用に保持しながら、新しいスキルベースのアプローチへ誘導する構成は移行期に有用です。
この設計判断を
.claude/memory/decisions.mdに記録することを推奨します。Based on learnings: "Document decisions and decisions in.claude/memory/decisions.md(Why) and reusable patterns in.claude/memory/patterns.md(How) as the Single Source of Truth (SSOT)"skills/harness-mem/SKILL.md (1)
1-186: Claude-mem統合スキルの定義が包括的で実用的です。OS検出、依存関係チェック（Bun、Claude-mem）、ハーネスモード設定、検証までの完全なワークフローが文書化されており、クロスセッション学習の具体例も価値提案を明確にしています。日本語ローカライゼーションオプションの提供も適切です。
静的解析ツールが指摘する軽微なフォーマット問題（Line 135のテーブル形式、Line 143のコードブロック言語指定）を修正すると、一貫性がさらに向上します：
🔧 フォーマット修正案
Line 135のテーブル形式を修正：
-| Event | Recorded Data | -|-------|---------------| +| Event | Recorded Data | +|-----------------|---------------------------------------|Line 143のコードブロックに言語を追加：
-``` +```text Session N: Error "User type not found"skills/setup-tools/references/opencode-setup.md (1)
37-37: 見出し代わりの強調表記は見出しに置換してください。
markdownlintの指摘通り、**Wait for response**は小見出し化した方が読みやすいです。skills/setup-tools/references/ci-setup.md (1)
35-53: 強調表記の“Wait for response”は見出し化が望ましいです。
同一パターンが複数あるため、#### Wait for response等に統一してください。skills/setup-tools/references/webhook-setup.md (1)
161-161: 裸URLはリンク化してください。
markdownlintの指摘通り、URLはリンク化した方が可読性が上がります。skills/harness-ui/SKILL.md (2)
80-85:openはmacOS限定なので他OS向けの案内を追加してください。
Linux/Windows利用者が詰まるため、xdg-open/start等の代替を併記するのが安全です。🛠 追記例
-# Open in browser -open http://localhost:3001 +# Open in browser +open http://localhost:3001 # macOS +# xdg-open http://localhost:3001 # Linux +# start http://localhost:3001 # Windows
59-59: 裸URLはリンク化してください。
markdownlintの指摘通り、URLはリンク化した方が可読性が上がります。skills/setup-tools/references/mcp-setup.md (1)
109-125: コードブロックに言語指定を追加してください。
markdownlintの指摘通り、例示ブロックにはtext等の言語指定が必要です。skills/harness-init/SKILL.md (1)
65-71: テーブルの書式について（オプション）静的解析ツールがテーブルの列の配置（パイプ記号の位置）について指摘していますが、これは表示に影響しない書式の問題です。必要に応じて整形できますが、機能には影響しません。
skills/harness-init/references/smart-defaults.md (1)
40-43: 2-Agentワークフローの説明を確認2-Agent モードの使用手順が明確に記載されています。Line 41の "Consult with Cursor" は文法的に正しいですが、より簡潔に "Consult Cursor" とすることも可能です（オプション）。
skills/planning/references/execution-flow.md (1)
46-54: コードブロックの言語指定を推奨WebSearchの使用例を示すコードブロック（lines 50-54）に言語指定子がありません。
textまたは適切な識別子を追加することで、可読性が向上します。♻️ 言語指定子の追加案
-``` +```text WebSearch for: - "{{project type}} tech stack 2025" - "{{similar service}} architecture"</details> </blockquote></details> <details> <summary>skills/work/SKILL.md (1)</summary><blockquote> `3-47`: **`argument-hint` と Options の不整合を解消してください。** `argument-hint` に `--isolation` / `--max-iterations` / `--skip-review` / `--reason` が含まれていないため、利用者が見落とす可能性があります。どちらかに揃えてください。 </blockquote></details> <details> <summary>skills/harness-update/references/verification.md (1)</summary><blockquote> `83-86`: **次のステップに `/harness-review` を追加するのを推奨します。** 品質チェックの導線が増えるため、更新後の確認フローが明確になります。 <details> <summary>➕ 追記例</summary> ```diff > **Next steps:** > - "`/sync-status`" → Check current status > - "`/plan-with-agent` I want to build XXX" → Add new tasks > - "`/work`" → Execute Plans.md tasks +> - "`/harness-review`" → Run quality checksBased on learnings: “Use
/plan-with-agentto add tasks to Plans.md,/workto implement tasks,/harness-reviewfor quality checks, and./tests/validate-plugin.shfor structure validation”.skills/memory/references/sync-project-specs.md (1)
48-50: コードブロックに言語指定を追加してください。マーカー状態遷移を示すコードフェンスに言語識別子がありません。可読性向上のため
textまたはplaintextを指定することを推奨します。📝 修正案
-``` +```text pm:依頼中 → cc:WIP → cc:完了 → pm:確認済</details> </blockquote></details> <details> <summary>skills/memory/references/sync-ssot-from-memory.md (2)</summary><blockquote> `49-59`: **コードブロックに言語指定を追加してください。** メモリ検索コマンド例を示すコードフェンスに言語識別子がありません。これらは実行コマンド例なので、`bash`または`shell`の指定を推奨します。 <details> <summary>📝 修正案</summary> ```diff -``` +```bash mem-search: type:decision mem-search: type:discovery concepts:pattern mem-search: type:bugfix concepts:gotchaFor Serena:
-+bash
mcp__serena__list_memories
mcp__serena__read_memory (target memories)
185-188: コードブロックに言語指定を追加してください。フォールバックメッセージ例を示すコードフェンスに言語識別子がありません。
textまたはplaintextの指定を推奨します。📝 修正案
-``` +```text > Cannot access memory system. > Please paste the information you want to promote.</details> </blockquote></details> <details> <summary>skills/ultrawork/references/security-guards.md (1)</summary><blockquote> `22-32`: **実験的機能の制限をより具体的に示すことを推奨します** 技術的制限が列挙されていますが、グロブや変数展開が評価されない場合の具体的なリスク例があると、ユーザーの理解がさらに深まります。 <details> <summary>💡 制限の具体例を追加する提案</summary> ```diff ### 技術的制限 | 制限 | 説明 | |------|------| | **展開前の文字列のみ** | シェル展開後の実パスは取得不可 | | **グロブ・変数未対応** | `$DIR/*` は正しく評価できない | | **rm 以外は対象外** | `find -delete` 等は検出不可 | + +**具体例**: +- `allowed_rm_paths: ["temp"]` でも `rm -rf $TEMP_DIR` は検出不可（変数展開前に評価） +- `rm -rf node_modules/*` のグロブ展開結果は検証されない
| - Run `/sync-status` to verify markers | ||
| - Use `/remember` to lock "project-specific operations" if needed |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Search for sync-status command/skill definition
echo "=== Searching for sync-status ==="
find . -type f \( -name "*.md" -o -name "*.ts" -o -name "*.js" -o -name "*.json" \) | head -200 | xargs rg -l "sync-status" 2>/dev/null | head -20
echo ""
echo "=== Searching for remember command/skill ==="
find . -type f \( -name "*.md" -o -name "*.ts" -o -name "*.js" -o -name "*.json" \) | head -200 | xargs rg -l "remember" 2>/dev/null | head -20Repository: Chachamaru127/claude-code-harness
Length of output: 862
🏁 Script executed:
#!/bin/bash
# List directory structure to understand codebase layout
echo "=== Repository structure ==="
ls -la
echo ""
echo "=== Skills directory structure ==="
find skills -type f -name "*.md" 2>/dev/null | head -20
echo ""
echo "=== Commands directory structure ==="
find commands -type f -name "*.md" 2>/dev/null | head -20Repository: Chachamaru127/claude-code-harness
Length of output: 3604
🏁 Script executed:
#!/bin/bash
# Check the specific file mentioned in the review
echo "=== Content of sync-project-specs.md lines 80-95 ==="
if [ -f "skills/memory/references/sync-project-specs.md" ]; then
sed -n '80,95p' "skills/memory/references/sync-project-specs.md"
else
echo "File not found at expected path, searching..."
find . -name "sync-project-specs.md" -type f
fiRepository: Chachamaru127/claude-code-harness
Length of output: 584
🏁 Script executed:
#!/bin/bash
# Search for remember command definition
echo "=== Searching for remember command definition ==="
find commands -type f -name "*remember*" 2>/dev/null
echo ""
echo "=== Checking if /remember is referenced as a command ==="
find commands -type f -name "*.md" | xargs rg "/remember" 2>/dev/null | head -10
echo ""
echo "=== Checking skills/memory structure ==="
find skills/memory -type f -name "*.md" 2>/dev/null | head -20Repository: Chachamaru127/claude-code-harness
Length of output: 576
🏁 Script executed:
#!/bin/bash
# Check the sync-status command definition to understand the pattern
echo "=== Content of sync-status command ==="
cat ./opencode/commands/core/sync-status.md 2>/dev/null || cat ./commands/core/sync-status.md 2>/dev/null || echo "Not found in expected paths"
echo ""
echo "=== Checking for remember in memory skill ==="
find skills/memory -type f -name "*.md" | xargs rg -l "/remember" 2>/dev/nullRepository: Chachamaru127/claude-code-harness
Length of output: 5111
/rememberコマンドが実装されていません。削除または実装してください。
検証結果:
/sync-status: ✅ 実装済み (commands/core/sync-status.mdに存在)/remember: ❌ 実装されていない (トラッキング項目#10350のみ存在)
87-88行目で/rememberコマンドの使用を推奨していますが、このコマンドはまだ実装されていません。ドキュメントの指示に従うユーザーが存在しないコマンドを実行しようとするため、以下のいずれかを実施してください:
/rememberコマンドを実装する- 指示から
/rememberへの言及を削除する /rememberを別の実装済みコマンドに置き換える
🤖 Prompt for AI Agents
In `@skills/memory/references/sync-project-specs.md` around lines 87 - 88, The doc
mentions a non-existent `/remember` command; either implement the `/remember`
command (resolve or link to tracking item `#10350` and add its command handler and
docs) or remove/replace its mention in the project spec text: edit the
"project-specific operations" guidance in the sync-project-specs content to
delete the `/remember` line or swap it for an existing command such as
`/sync-status` (and update the surrounding wording accordingly) so the document
no longer instructs users to run an unimplemented `/remember`.
| run: | | ||
| CHANGED_FILES=$(git diff --name-only origin/${{ github.base_ref }}...HEAD) | ||
| claude --non-interactive << 'EOF' | ||
| /harness-review --ci --files "$CHANGED_FILES" | ||
| EOF |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# Find the webhook setup file
find . -name "webhook-setup.md" -type fRepository: Chachamaru127/claude-code-harness
Length of output: 171
🏁 Script executed:
# Check if file exists and display lines 67-71
if [ -f "skills/setup-tools/references/webhook-setup.md" ]; then
echo "=== Lines 67-71 of webhook-setup.md ==="
sed -n '67,71p' "skills/setup-tools/references/webhook-setup.md"
echo ""
echo "=== Context (lines 60-75) ==="
sed -n '60,75p' "skills/setup-tools/references/webhook-setup.md"
else
echo "File not found at expected location"
find . -type f -name "*webhook*" 2>/dev/null | head -20
fiRepository: Chachamaru127/claude-code-harness
Length of output: 861
🏁 Script executed:
# Create a test script to verify bash behavior with here-docs
python3 << 'PYTHON_EOF'
import subprocess
test_cases = [
{
"name": "Single-quoted EOF (variable expansion disabled)",
"script": """
CHANGED_FILES="file1.txt file2.txt"
claude_output=$(cat << 'EOF'
/harness-review --files "$CHANGED_FILES"
EOF
)
echo "$claude_output"
"""
},
{
"name": "Unquoted EOF (variable expansion enabled)",
"script": """
CHANGED_FILES="file1.txt file2.txt"
claude_output=$(cat << EOF
/harness-review --files "$CHANGED_FILES"
EOF
)
echo "$claude_output"
"""
}
]
for test in test_cases:
print(f"\n{test['name']}:")
print("-" * 50)
result = subprocess.run(['bash', '-c', test['script']], capture_output=True, text=True)
print(result.stdout)
if result.stderr:
print(f"stderr: {result.stderr}")
PYTHON_EOFRepository: Chachamaru127/claude-code-harness
Length of output: 361
CHANGED_FILESが展開されず、実際の差分ファイルが渡りません。
<< 'EOF'のシングルクォート引用は変数展開を無効化するため、$CHANGED_FILESが文字列リテラルのまま/harness-reviewコマンドに渡されます。その結果、実際のファイル一覧の代わりに"$CHANGED_FILES"という文字列が引数として使用されるため、ファイル変更検出機能が機能不全に陥ります。
修正案
- claude --non-interactive << 'EOF'
- /harness-review --ci --files "$CHANGED_FILES"
- EOF
+ claude --non-interactive << EOF
+ /harness-review --ci --files "$CHANGED_FILES"
+ EOF📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| run: | | |
| CHANGED_FILES=$(git diff --name-only origin/${{ github.base_ref }}...HEAD) | |
| claude --non-interactive << 'EOF' | |
| /harness-review --ci --files "$CHANGED_FILES" | |
| EOF | |
| run: | | |
| CHANGED_FILES=$(git diff --name-only origin/${{ github.base_ref }}...HEAD) | |
| claude --non-interactive << EOF | |
| /harness-review --ci --files "$CHANGED_FILES" | |
| EOF |
🤖 Prompt for AI Agents
In `@skills/setup-tools/references/webhook-setup.md` around lines 67 - 71,
現在のhere-doc uses single-quoted delimiter << 'EOF' which prevents shell variable
expansion so CHANGED_FILES is passed literally; change the heredoc to an
unquoted delimiter (<<EOF) so $CHANGED_FILES expands, and ensure the command
inside uses proper quoting like /harness-review --ci --files "$CHANGED_FILES"
(refer to the CHANGED_FILES assignment and the heredoc block using << 'EOF' in
the snippet) to pass the actual file list to the harness-review command.
- Remove deprecated /cc-cursor-cc references from plan-with-cc templates - Rename dev-browser to agent-browser in docs - Sync opencode/ with new skill structure Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove old skill directories (dev-browser, docs, video, workflow) - Update CLAUDE.md with new skill names (handoff, notebookLM, generate-video, agent-browser) - Update all documentation references - Sync opencode/ with updated structure - Fix video-scene-generator agent references Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Fix all issues with AI agents
In `@agents/video-scene-generator.md`:
- Around line 32-33: The two referenced docs listed in
agents/video-scene-generator.md—`.claude/skills/generate-video/references/quality-patterns.md`
and `.claude/skills/generate-video/references/best-practices.md`—do not exist;
either add or move the actual markdown files into that exact
`.claude/skills/generate-video/references/` directory, or update the links in
agents/video-scene-generator.md to point to the correct existing paths (use the
real filenames/locations in the repo). Ensure the final paths resolve in the
repository so the agent can load the referenced documents at runtime.
In `@docs/HARNESS_COMPLETE_MAP.md`:
- Around line 509-510: The Markdown table rows for the entries `notebookLM` and
`maintenance` have misaligned pipe columns triggering MD060; fix by aligning the
pipe characters for all columns so each row matches the same column widths (pad
the cells with spaces so the pipes line up vertically), e.g. adjust the two
lines containing `| `notebookLM` | ... |` and `| `maintenance` | ... |` so
their cell contents have consistent spacing and the table pipes are vertically
aligned.
- Around line 491-492: The Markdown table rows for `handoff` and
`workflow-guide` are misaligned and triggering MD060; fix by aligning the pipe
characters vertically so each column has matching pipe positions across all
lines (adjust the two rows to match the header/other rows), e.g., ensure the
first column uses backticked keys `handoff` and `workflow-guide`, the second
column contains the Japanese descriptions, and the third column contains the
quoted tags, with pipes placed consistently for `handoff` and `workflow-guide`
to match the rest of the table.
| > - [skills/generate-video/references/quality-patterns.md](../.claude/skills/generate-video/references/quality-patterns.md) - V8品質基準 | ||
| > - [skills/generate-video/references/best-practices.md](../.claude/skills/generate-video/references/best-practices.md) - SaaS動画ガイドライン |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify the new documentation paths exist
echo "=== Checking for new generate-video reference files ==="
find . -path "*/.claude/skills/generate-video/references/quality-patterns.md" -o -path "*/.claude/skills/generate-video/references/best-practices.md" 2>/dev/null | head -20
echo ""
echo "=== Listing directory structure under .claude/skills/generate-video/ ==="
find ./.claude/skills/generate-video -type f 2>/dev/null | head -20
echo ""
echo "=== Checking for remaining old 'video' path references in agents directory ==="
rg -n "\.claude/skills/video/references" agents/ 2>/dev/null || echo "No old path references found in agents/"
echo ""
echo "=== Checking for any 'video' skill references in the specific file ==="
cat agents/video-scene-generator.md 2>/dev/null | head -40Repository: Chachamaru127/claude-code-harness
Length of output: 1461
参照先のドキュメントファイルが存在しません。
Lines 32-33で参照されている以下のファイルがリポジトリ内に見つかりません：
.claude/skills/generate-video/references/quality-patterns.md.claude/skills/generate-video/references/best-practices.md
スキル名の変更後、これらのドキュメントファイルを正しい位置に配置するか、または存在するパスに参照を修正してください。現在のままでは、このエージェントは起動時に必須のドキュメント参照にアクセスできません。
🧰 Tools
🪛 markdownlint-cli2 (0.20.0)
[warning] 32-32: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 32-32: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 32-32: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 32-32: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
🤖 Prompt for AI Agents
In `@agents/video-scene-generator.md` around lines 32 - 33, The two referenced
docs listed in
agents/video-scene-generator.md—`.claude/skills/generate-video/references/quality-patterns.md`
and `.claude/skills/generate-video/references/best-practices.md`—do not exist;
either add or move the actual markdown files into that exact
`.claude/skills/generate-video/references/` directory, or update the links in
agents/video-scene-generator.md to point to the correct existing paths (use the
real filenames/locations in the repo). Ensure the final paths resolve in the
repository so the agent can load the referenced documents at runtime.
| | `handoff` | ハンドオフ管理、レビューコメントの自動修正 | 「ハンドオフ」「PMに報告」「自動修正」 | | ||
| | `workflow-guide` | Cursor ↔ Claude Code 2-Agent ワークフローの解説 | 「ワークフロー」「コラボ」「プロセス」 | |
There was a problem hiding this comment.
表の列揃え（MD060）を修正してください。
markdownlint の警告対象になっているため、パイプ位置を揃えてください。
🧰 Tools
🪛 markdownlint-cli2 (0.20.0)
[warning] 491-491: Table column style
Table pipe has extra space to the left for style "compact"
(MD060, table-column-style)
[warning] 491-491: Table column style
Table pipe has extra space to the left for style "compact"
(MD060, table-column-style)
[warning] 492-492: Table column style
Table pipe has extra space to the left for style "compact"
(MD060, table-column-style)
🤖 Prompt for AI Agents
In `@docs/HARNESS_COMPLETE_MAP.md` around lines 491 - 492, The Markdown table rows
for `handoff` and `workflow-guide` are misaligned and triggering MD060; fix by
aligning the pipe characters vertically so each column has matching pipe
positions across all lines (adjust the two rows to match the header/other rows),
e.g., ensure the first column uses backticked keys `handoff` and
`workflow-guide`, the second column contains the Japanese descriptions, and the
third column contains the quoted tags, with pipes placed consistently for
`handoff` and `workflow-guide` to match the rest of the table.
| | `notebookLM` | NotebookLM YAML やスライド用ドキュメント生成 | 「ドキュメント」「NotebookLM」「スライド」 | | ||
| | `maintenance` | プロジェクトファイルの整理・クリーンアップ | 「整理して」「クリーンアップ」 | |
There was a problem hiding this comment.
表の列揃え（MD060）を修正してください。
markdownlint の警告対象になっているため、パイプ位置を揃えてください。
🧰 Tools
🪛 markdownlint-cli2 (0.20.0)
[warning] 509-509: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 509-509: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 509-509: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 510-510: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 510-510: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
[warning] 510-510: Table column style
Table pipe does not align with header for style "aligned"
(MD060, table-column-style)
🤖 Prompt for AI Agents
In `@docs/HARNESS_COMPLETE_MAP.md` around lines 509 - 510, The Markdown table rows
for the entries `notebookLM` and `maintenance` have misaligned pipe columns
triggering MD060; fix by aligning the pipe characters for all columns so each
row matches the same column widths (pad the cells with spaces so the pipes line
up vertically), e.g. adjust the two lines containing `| `notebookLM` | ... |`
and `| `maintenance` | ... |` so their cell contents have consistent spacing and
the table pipes are vertically aligned.
- Change docs to notebookLM in README template Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Accept PR #36 versions for skill files (include proper skill renaming) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2 participants
🎯 What's Changed for You
スキル名の明確化と内部スキルの整理
Before → After
Changed
🤖 Generated with Claude Code
Summary by CodeRabbit
リリースノート
新機能
改善
その他
✏️ Tip: You can customize this high-level summary in your review settings.