Skip to content

利用claude code agent框架一步一步实现deep research!很强大很简单的skills。我一步一步介绍实现deep research,因为deep research就是agent框架第一应用,对比一下各个框架实现这个deep research,就知道哪个框架才是真厉害。

Notifications You must be signed in to change notification settings

liangdabiao/Claude-Code-Deep-Research-main

Repository files navigation

Claude Code Deep Research Agent

A sophisticated multi-agent research framework that implements OpenAI's Deep Research and Google Gemini's Deep Research capabilities using Claude Code's native features.

Overview

This project leverages Claude Code's Skills and Commands system to conduct comprehensive, citation-backed research through:

  • Graph of Thoughts (GoT) Framework - Intelligent research path management with graph-based reasoning
  • 7-Phase Deep Research Process - Structured methodology for quality research
  • Multi-Agent Architecture - Parallel research agents with specialized roles
  • Citation Validation System - A-E source quality ratings with chain-of-verification

Quick Start

Prerequisites

  • Claude Code CLI installed
  • Active Claude Code account with API access

Installation

  1. Clone this repository:
git clone <repository-url>
cd Claude-Code-Deep-Research-main
  1. The Skills and Commands are already configured in .claude/ directory

Basic Usage

The simplest way to conduct deep research:

/deep-research [your research topic]

Example:

/deep-research AI applications in clinical diagnosis

This single command will:

  1. Ask clarifying questions to refine your research needs
  2. Create a structured research plan
  3. Deploy multiple parallel research agents
  4. Synthesize findings into a comprehensive report
  5. Validate all citations
  6. Output results to RESEARCH/[topic]/ directory

Advanced Usage

Step-by-Step Research Workflow

For more control over the research process:

1. Refine Your Question

/refine-question Should I use WebAssembly for my project?

The Question Refiner will ask 5-6 clarifying questions about:

  • Specific focus areas
  • Output format requirements
  • Geographic and time scope
  • Target audience
  • Special requirements

2. Plan Research (Optional)

/plan-research [structured prompt from step 1]

Creates a detailed execution plan showing:

  • How the topic breaks into subtopics
  • Which agents will be deployed
  • Expected timeline

3. Execute Research

/deep-research [your topic]

4. Synthesize Findings (If needed)

/synthesize-findings RESEARCH/[topic]/research_notes/

5. Validate Citations

/validate-citations RESEARCH/[topic]/full_report.md

Project Structure

claude-code-deep-research/
├── .claude/
│   ├── skills/                    # Research skills
│   │   ├── question-refiner/      # Question refinement
│   │   ├── research-executor/     # Main research execution
│   │   ├── got-controller/        # Graph of Thoughts controller
│   │   ├── citation-validator/    # Citation validation
│   │   └── synthesizer/           # Research synthesis
│   ├── commands/                  # User commands
│   │   ├── deep-research.md       # Main research command
│   │   ├── refine-question.md     # Question refinement
│   │   ├── plan-research.md       # Research planning
│   │   ├── synthesize-findings.md # Findings synthesis
│   │   └── validate-citations.md  # Citation validation
│   └── settings.local.json        # Tool permissions
├── RESEARCH/                      # Research outputs
│   └── [topic_name]/
│       ├── README.md
│       ├── executive_summary.md
│       ├── full_report.md
│       ├── data/
│       ├── visuals/
│       ├── sources/
│       ├── research_notes/
│       └── appendices/
├── CLAUDE.md                      # Quick reference for Claude Code
├── CLAUDE2.md                     # Graph of Thoughts guide
├── PROJECT_UNDERSTANDING.md       # Architecture documentation
├── IMPLEMENTATION_GUIDE.md        # User guide with examples
└── README.md                      # This file

Research Output Structure

Each research project creates a structured output:

RESEARCH/[topic_name]/
├── README.md                    # Overview and navigation
├── executive_summary.md         # 1-2 page key findings
├── full_report.md               # Complete analysis (20-50 pages)
├── data/
│   └── statistics.md            # Key numbers and facts
├── visuals/
│   └── descriptions.md          # Chart/graph descriptions
├── sources/
│   ├── bibliography.md          # Complete citations
│   └── source_quality_table.md  # A-E quality ratings
├── research_notes/
│   └── agent_findings_summary.md # Raw agent outputs
└── appendices/
    ├── methodology.md           # Research methods
    └── limitations.md           # Unknowns and gaps

Citation Requirements

Every factual claim includes:

  1. Author/Organization name
  2. Publication date
  3. Source title
  4. Direct URL/DOI
  5. Page numbers (if applicable)

Source Quality Ratings:

  • A: Peer-reviewed RCTs, systematic reviews, meta-analyses
  • B: Cohort studies, clinical guidelines, reputable analysts
  • C: Expert opinion, case reports, mechanistic studies
  • D: Preprints, preliminary research, blogs
  • E: Anecdotal, theoretical, speculative

Graph of Thoughts Framework

The GoT framework manages research as a graph with these operations:

Operation Purpose Example
Generate(k) Spawn k parallel research paths Generate(4) from root → 4 research paths
Aggregate(k) Merge k findings into synthesis Aggregate(3) → 1 comprehensive report
Refine(1) Improve existing finding Refine(node_5) → Enhanced quality
Score Rate quality (0-10) Score based on citations, accuracy
KeepBestN(n) Prune to top n nodes KeepBestN(3) → Retain best 3

Research Patterns:

  • Balanced: Generate(4-5) → Score best → Deepen top → Aggregate
  • Depth-first: Generate(3) → Take best → Generate(3) from it
  • Breadth-first: Generate(8) → KeepBestN(5) → Generate(2) each

Documentation

Document Description
CLAUDE.md Quick reference for Claude Code instances
CLAUDE2.md Complete Graph of Thoughts implementation
PROJECT_UNDERSTANDING.md Detailed architecture and design
IMPLEMENTATION_GUIDE.md User guide with examples and workflows

Commands Reference

Command Usage Description
/deep-research /deep-research [topic] Execute complete research workflow
/refine-question /refine-question [question] Refine into structured prompt
/plan-research /plan-research [prompt] Create execution plan
/synthesize-findings /synthesize-findings [dir] Combine research outputs
/validate-citations /validate-citations [file] Verify citation quality

Examples

Market Research

/deep-research AI in healthcare market, focus on clinical diagnosis,
             comprehensive report, global scope, 2022-2024 data,
             audience is healthcare executives

Technical Assessment

/deep-research WebAssembly vs JavaScript performance benchmarks

Academic Literature Review

/deep-research Transformer architectures in AI,
             peer-reviewed sources only, 2017-present,
             comprehensive literature review

Features

  • Multi-agent parallel research (3-8 agents simultaneously)
  • Graph of Thoughts optimization for quality
  • Automatic citation validation
  • Source quality ratings (A-E scale)
  • Chain-of-verification to prevent hallucinations
  • Structured output with executive summaries
  • Cross-source triangulation

Performance

  • Quick research (narrow topic): 10-15 minutes
  • Standard research (moderate scope): 20-30 minutes
  • Comprehensive research (broad scope): 30-60 minutes
  • Academic literature review: 45-90 minutes

Contributing

Contributions are welcome! To add new skills or improvements:

  1. Follow the skill structure in .claude/skills/
  2. Include SKILL.md, instructions.md, examples.md
  3. Test with diverse research topics
  4. Update documentation

License

This project is provided as-is for educational and research purposes.

Acknowledgments

  • Graph of Thoughts framework inspired by SPCL, ETH Zürich
  • Built with Claude Code
  • 7-Phase Research Process based on deep research best practices

For detailed usage instructions, see IMPLEMENTATION_GUIDE.md

For architecture details, see PROJECT_UNDERSTANDING.md

About

利用claude code agent框架一步一步实现deep research!很强大很简单的skills。我一步一步介绍实现deep research,因为deep research就是agent框架第一应用,对比一下各个框架实现这个deep research,就知道哪个框架才是真厉害。

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published