Claude Research Framework

한국어 안내 Claude Code에서 체계적인 리서치와 AI 네이티브 학습을 도와주는 플러그인 모음입니다.

포함된 플러그인:

• domain-research: 대화형 리서치 파이프라인 (어떤 분야든 적용 가능)
• agentic-learning: Claude Code 에이전틱 학습 커리큘럼 (Rev 0~6, 7개 레볼루션)
• pdf-research: PDF 문서 인덱싱 및 시맨틱 검색 (LightRAG 기반)
• pm-coach: PM 업무 소통 최적화

---

A collection of research plugins for Claude Code that help you conduct systematic research through conversational analysis and semantic document search.

Plugins

---

Quick Start

Installation

# Add this marketplace to Claude Code
/plugin marketplace add hongsw/plugin-for-claude-research

# Install plugins
/plugin install domain-research
/plugin install agentic-learning
/plugin install pdf-research
/plugin install pm-coach

Usage

Domain Research:

You: "I'm interested in AI for healthcare"
Claude: [Conversational discovery → Research context → 5-step pipeline]

Agentic Learning:

You: /agentic-learning
Claude: [레볼루션 0부터 시작 → 환경 설정 → CLI 이해 → 핵심 기능 학습]

You: /learn rev3
Claude: [Rev 3 핵심 기능 레볼루션으로 바로 이동]

PDF Research:

You: /pdf-research ~/Documents/papers 인덱싱해줘
Claude: [Indexes PDFs → Enables semantic search]

You: AI 인재 전략에 대해 알려줘
Claude: [Searches indexed documents → Returns relevant insights]

---

Screenshots

PDF Research 워크플로우

Screen 3: PDF Research 상태 확인

• /pdf-research 명령어로 스킬 실행
• 현재 설정 상태 확인 (PDF 디렉토리, 스토리지, 검색 모드)
• 사용 가능한 명령어 안내 (Index, Search, Configure)

Screen 4: PDF 인덱싱 및 검색

• PDF 폴더 지정 후 자동 인덱싱
• LightRAG 기반 시맨틱 검색 실행
• 검색 결과에서 관련 문서 내용 추출

---

Plugin Details

1. Domain Research

Universal research framework that guides users from broad exploration to specific domain research.

Features:

• Conversational intent discovery
• 5-step research pipeline
• MCP integration (WebSearch, Sequential)
• Works for any domain

Pipeline:

1. Conversational Intent Analysis
2. Key Question Generation
3. Research Gap Identification
4. Multi-Source Synthesis
5. Practical Application

2. PDF Research

LightRAG-based semantic search over PDF documents.

Features:

• PDF text extraction and chunking
• Knowledge graph construction
• Vector embeddings (OpenAI)
• Multiple search modes (hybrid, local, global, naive)

Commands:

# Configure
python pdf_research.py config --pdf-dir /path/to/pdfs

# Index
python pdf_research.py index

# Search
python pdf_research.py search "your question"

# Status
python pdf_research.py status

Requirements:

• Python 3.10+
• OpenAI API key
• Dependencies: lightrag-hku[api], pymupdf, python-dotenv

2. Agentic Learning

AI-native coding curriculum that teaches Claude Code through hands-on agentic workflows.

Curriculum (7 Revolutions):

Commands:

/agentic-learning        # 처음부터 시작
/learn rev3             # 특정 레볼루션으로 이동
/learn skill            # 스킬 만들기 실습
/learn research         # 리서치 결합 학습

---

3. PM Coach

PM communication optimizer for Korean users.

Modes:

• Request mode: 요청 최적화
• Receive mode: 수신 정리
• Report mode: 보고서 작성

---

Directory Structure

plugin-for-claude-research/
├── .claude-plugin/
│   ├── plugin.json              # Marketplace metadata
│   └── marketplace.json         # Plugin registry
├── plugins/
│   ├── domain-research/         # Research pipeline plugin
│   │   ├── skills/
│   │   │   ├── domain-research/ # Core research skill
│   │   │   └── agentic-learning/# AI-native learning curriculum
│   │   │       ├── SKILL.md
│   │   │       ├── prompts/     # 7 revolution prompt files
│   │   │       └── references/  # Rev 0~6 reference materials
│   │   └── bin/install.js
│   ├── pdf-research/            # PDF semantic search plugin
│   │   ├── skills/pdf-research/
│   │   │   ├── SKILL.md
│   │   │   ├── prompts/
│   │   │   └── scripts/         # Python CLI tools
│   │   └── bin/install.js
│   └── pm-coach/                # PM communication plugin
│       └── commands/
├── templates/                   # Research input templates
├── CHANGELOG.md
└── README.md

---

Domain Research Pipeline

Step 0: Conversational Intent Analysis

Start with dialogue, not forms. Adaptive questioning based on user clarity.

Step 1: Key Question Generation

Generate 5 testable, meaningful research questions.

Step 2: Research Gap Identification

Identify underexplored areas and emerging opportunities.

Step 3: Key Insight Extraction

Deep analysis of individual sources.

Step 4: Multi-Source Synthesis

Integrate findings across sources.

Step 5: Practical Application

Transform insights to actionable items.

---

PDF Research Search Modes

---

User Types Supported

“I know exactly what I want”

You: "I want to research computer vision for quality inspection"
→ 2-3 clarifying questions → Research context ready

“I have a vague idea”

You: "I'm interested in AI for my factory"
→ Discovery questions → Focused direction → Context ready

“I want to search my documents”

You: /pdf-research ~/papers 인덱싱해줘
→ PDF indexed → Semantic search enabled

---

MCP Server Integration

---

Requirements

Domain Research

• Claude Code CLI

PDF Research

• Python 3.10+
• OpenAI API key (OPENAI_API_KEY)
• Dependencies:

  pip install lightrag-hku[api] pymupdf python-dotenv
  

---

Contributing

Contributions welcome! Areas of interest:

• New domain examples
• Prompt refinements
• Additional plugins
• Language support

---

License

MIT License - see LICENSE file for details.

---

Acknowledgments

• Inspired by plugins-for-claude-natives
• PDF search powered by LightRAG