Menu
リサーチから Skill/Subagent 作成までを1コマンドで実行するパイプライン。トピックを指定すると、Webリサーチ → ベストプラクティス抽出 → Skill/Subagent生成 → バリデーションまで自動実行。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Analyzes repository structure and generates CLAUDE.md documentation. Use when user mentions "onboard", "analyze project", "understand codebase", "project structure", or when working with new/unfamiliar repositories.
Automatically resume previous work by searching claude-mem for incomplete tasks, past decisions, and session context. Use when starting a session, continuing work, or when user says "resume", "continue", "前回", "再開", "続き", "what was I working on", or asks about previous work.
Claude Code Subagentsのレビューと品質改善。ベストプラクティスに基づいてSubagentを分析し、評価レポートと改善提案を提供。"review agent", "agentをレビュー", "subagent品質チェック" で起動。
Create, edit, or audit Claude Code custom agents (subagents), including file locations, YAML frontmatter, tool/permission settings, skills preloading, and prompt design. Use when asked to define a new Claude Code subagent, update an existing one, or review subagent configs for correctness and best practices. Triggers on "create agent", "new subagent", "define custom agent", "audit agent".
Comprehensive .claude/ directory inspector. Orchestrates multiple specialized subagents to analyze settings, hooks, memory, MCP, rules, and skills/subagents classification. Provides prioritized recommendations with examples and rationale.
Claude Code Skillsのレビューと品質改善。ベストプラクティスに基づいてSkillを分析し、評価レポートと改善提案を提供。"review skill", "skillをレビュー", "skill品質チェック" で起動。
| name | skill-pipeline |
| description | リサーチから Skill/Subagent 作成までを1コマンドで実行するパイプライン。トピックを指定すると、Webリサーチ → ベストプラクティス抽出 → Skill/Subagent生成 → バリデーションまで自動実行。 |
| allowed-tools | ["Task","Read","Write","Bash"] |
| model | sonnet |
リサーチから実装までを一気通貫で実行するパイプライン。
Required subagents (.claude/agents/):
parallel-researcher.md - Stage 1: Researchauto-validator.md - Stage 4: ValidateOptional subagents:
skill-reviewer.md - Optional quality checkError if missing: Pipeline will report which subagent is unavailable and suggest creating it.
/skill-pipeline [topic] [--type skill|subagent]
例:
/skill-pipeline "Go error handling best practices" --type skill/skill-pipeline "TypeScript型安全性" --type subagentTask: subagent_type=parallel-researcher
Prompt: "[topic]についてベストプラクティスを調査。Skill/Subagent作成に必要な情報を収集"
Output: .research/[topic]/synthesis.json
リサーチ結果から Skill/Subagent の設計を生成:
name: [derived from topic]
description: [1-line summary]
type: skill | subagent
triggers: [when to use]
core_functionality:
- [capability 1]
- [capability 2]
references_needed:
- [reference doc 1]
tools_required:
- [tool 1]
- [tool 2]
.claude/skills/[name]/
├── SKILL.md # メインスキル定義
├── references/ # 参考ドキュメント
│ └── best-practices.md
└── scripts/ # 必要に応じてスクリプト
└── validate.sh
.claude/agents/[name].md # サブエージェント定義
Task: subagent_type=auto-validator
Prompt: "生成した Skill/Subagent を検証: 構文チェック、必須フィールド確認、ベストプラクティス適合"
最終レポートを出力:
{
"pipeline_id": "skill-pipeline-[timestamp]",
"topic": "Original topic",
"type": "skill|subagent",
"research_summary": "Key findings from research",
"generated_files": [
".claude/skills/[name]/SKILL.md",
".claude/skills/[name]/references/best-practices.md"
],
"validation_result": {
"status": "pass|fail",
"issues": []
},
"next_steps": [
"テストしてみる: /[skill-name]",
"改善点があれば: /skill-pipeline feedback"
]
}
┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Research │───▶│ Design │───▶│ Generate │───▶│ Validate │───▶│ Report │
│ (parallel) │ │ │ │ │ │ (auto) │ │ │
└─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘
│ │ │ │ │
▼ ▼ ▼ ▼ ▼
.research/ design.yaml .claude/skills/ validation.json report.json
synthesis.json or agents/
| Stage | Failure | Recovery |
|---|---|---|
| Research | WebSearch fails | Retry with alternative queries |
| Design | Ambiguous requirements | Ask user for clarification |
| Generate | Missing template | Use default template |
| Validate | Syntax errors | Auto-fix and retry |
| Option | Description | Default | Status |
|---|---|---|---|
--type | skill or subagent | Inferred | Implemented |
--skip-research | Use existing research | false | Implemented |
--no-validate | Skip validation | false | Implemented |
--output-dir | Custom output directory | .claude/ | Implemented |
.research/[topic]/synthesis.json exists and --skip-research not set, ask to reuseUser: /skill-pipeline "Claude Code hooks automation"
Pipeline Starting...
[Stage 1/5] Research
├─ Spawning parallel-researcher...
├─ 4 research agents deployed
└─ Waiting for completion...
[Stage 2/5] Design
├─ Analyzing synthesis.json
├─ Type: skill (detected from topic)
└─ Name: hooks-automation
[Stage 3/5] Generate
├─ Creating .claude/skills/hooks-automation/
├─ Writing SKILL.md
└─ Writing references/hooks-patterns.md
[Stage 4/5] Validate
├─ Spawning auto-validator...
├─ Checking syntax: ✓
├─ Checking required fields: ✓
└─ Checking best practices: ✓
[Stage 5/5] Report
✓ Pipeline completed successfully
Generated:
- .claude/skills/hooks-automation/SKILL.md
- .claude/skills/hooks-automation/references/hooks-patterns.md
Try it: /hooks-automation
This skill orchestrates:
parallel-researcher (subagent) - For comprehensive researchauto-validator (subagent) - For validationskill-reviewer (subagent) - Optional quality checkFrom Claude Code official guidelines:
allowed-tools or toolsmodel: haiku for simple skills