research-to-diagram

Research to Diagram

Deeply research a topic and automatically generate a knowledge relationship graph PDF. A one-stop tool from research to visualization.

Description

This Skill receives a research topic provided by the user, automatically conducts deep web research, information collection, knowledge organization, and structural design, and ultimately generates a professional visualized relationship graph PDF. Unlike structure-to-pdf , this Skill focuses onactive research and knowledge mining, rather than passive data conversion.

Core Features

• Automatic Research: Use WebSearch for multiple rounds of deep research
• Intelligent Organization: Automatically extract, classify, and structure information
• Professional Design: Select the best visualization solution based on the characteristics of the topic
• Multiple Outputs: Supports tools such as Graphviz, PlantUML, Mermaid, etc.
• High-quality PDF: Generates vector graphics that can be scaled infinitely

Trigger Conditions

When the user wants to:

• Research the knowledge structure of a complex topic (such as "Dream of the Red Chamber character relationships")
• Generate character relationship diagrams, concept maps, knowledge graphs
• Understand the organizational structure or technical architecture of a certain field
• Visualize complex relationship networks
• Need to research and visualize a topic from scratch

Keywords：

• "research... and make a diagram"
• "study the relationships of..."
• "deeply analyze... and visualize"
• "generate... knowledge graph"
• "draw... relationship diagram" (no existing data)

Workflow

1. Task Planning (TodoWrite)

- 深度调研主题和相关知识
- 设计图谱结构和层次
- 创建可视化图表
- 生成PDF文档

2. Deep Research Stage

• Multiple Rounds of WebSearch: Collect information from different angles
  • Topic overview and background
  • Core elements and people/concepts
  • Relationships and connections
  • Hierarchies and classifications
• Information Source Recording: Save all reference links
• Knowledge Extraction: Identify key entities and relationships

3. Structural Design Stage

Select the best structure based on the topic type:

Character Relationship Diagram：

• Family lineage: multi-level tree structure
• Social network: network relationship diagram
• Organizational structure: hierarchical layout

Concept Map：

• Knowledge classification: tree diagram or mind map
• Concept relationships: directed graph
• Flowchart: linear or branching process

Technical Architecture：

• System components: modular layout
• Dependencies: hierarchical or network diagram
• Data flow: flowchart

4. Visualization Implementation

Prefer to use Graphviz (DOT language):

digraph G {
    // 全局样式
    graph [rankdir=TB bgcolor="#fdfdf5" fontname="Arial Unicode MS"]
    node [shape=box style="rounded,filled" fillcolor="#e8f4f8"]
    edge [fontname="Arial Unicode MS"]

    // 使用 subgraph cluster 分组
    subgraph cluster_group1 {
        label="分组名称"
        node1 [label="节点1"]
        node2 [label="节点2"]
    }

    // 定义关系
    node1 -> node2 [label="关系类型" color=red]
}

Alternative Tools：

• PlantUML: UML diagrams, sequence diagrams
• Mermaid: simple flowcharts, sequence diagrams

5. PDF Generation

dot -Tpdf diagram.dot -o output.pdf

6. Document Organization

Optionally generate documentation, including:

• Research topic overview
• Graph Legend
• Sources
• Usage Instructions

Design Principles

Visual Design

1. Color Coding: Use different colors to distinguish categories
2. Shape Distinction: Use different shapes for different types of entities
3. Clear Hierarchy: Use subgraph cluster for grouping
4. Relationship Annotation: Edge colors, styles, and labels express relationship types
5. Chinese Support: Use "Arial Unicode MS" or a system Chinese font

Information Hierarchy

1. Title Level: Main title
2. Group Level: Main categories/families
3. Entity Level: Specific people/concepts
4. Relationship Level: Connections and annotations
5. Legend Level: Explain the meaning of symbols

Layout Strategy

• rankdir=TB: Top-down (family trees, organizational charts)
• rankdir=LR: Left-to-right (flowcharts, timelines)
• rankdir=BT: Bottom-up (dependency graphs)
• splines=ortho: Orthogonal edges (clear organizational charts)
• splines=curved: Curved edges (aesthetically pleasing relationship networks)

Output Files

Default save location:~/Downloads/ or a user-specified directory

Generated files:

• <topic>_relations.dot - Graphviz source file
• <topic>_relations.pdf - Final PDF graph
• <topic>_sources.md - Reference materials (optional)

Usage Examples

Example 1: Character Relationships in a Literary Work

用户：深度调查《三国演义》里人物之间的关系，然后做个结构图 PDF

Skill Execution:

1. Research the main characters, factions, and relationships in Romance of the Three Kingdoms
2. Design: The three major factions of Wei, Shu, and Wu + character hierarchy + alliance/conflict relationships
3. Use Graphviz to create a multi-level relationship graph
4. Generate PDF

Example 2: Technical Concept Graph

用户：研究 Kubernetes 架构并生成可视化图谱

Skill Execution:

1. Research K8s core components and architecture layers
2. Design: Control plane/data plane/plugin ecosystem
3. Generate technical architecture diagram
4. Output PDF

Example 3: Historical Event Relationships

用户：分析二战主要国家和联盟关系，做成图表

Skill Execution:

1. Research participating countries, factions, and key time points
2. Design: Axis powers/Allied powers/neutral countries relationship network
3. Generate a relationship graph with a timeline
4. Output PDF

Differences from structure-to-pdf

Feature	research-to-diagram	structure-to-pdf
Input	Topic/research question only	Ready-made structured data
Research	✅ Automatic in-depth research	❌ No research needed
Knowledge Organization	✅ Automatic extraction and structuring	❌ Directly use user data
Use Cases	Knowledge exploration, research visualization	Fast data conversion
Time	Longer (requires research)	Fast

When to Use This Skill：

• ✅ There is no ready-made data and you need to research from scratch
• ✅ You want an in-depth understanding of the knowledge structure of a certain topic
• ✅ You need visualization supported by authoritative sources

When to use structure-to-pdf：

• ✅ Already have structured data
• ✅ Need quick conversion to diagrams
• ✅ Data format is simple and clear

Dependencies

Required：

• Graphviz: brew install graphviz (macOS)
• WebSearch: Built into Claude Code

Optional：

• PlantUML: brew install plantuml
• Mermaid CLI: npm install -g @mermaid-js/mermaid-cli

Configuration Options

Users can customize via parameters:

• --tool: Specify visualization tool (graphviz/plantuml/mermaid)
• --layout: Specify layout direction (TB/LR/BT/RL)
• --output: Specify output directory
• --depth: Research depth (quick/medium/deep)
• --sources: Whether to generate a reference document

Best Practices

1. Define the topic scope clearly: The more specific the topic, the clearer the diagram
2. Group reasonably: Use cluster to group related entities
3. Control complexity: A single diagram should not exceed 50 nodes
4. Refine progressively: Generate an overview diagram first, then go into details
5. Consistent colors: Use the same color scheme for entities of the same type
6. Cite sources: Cite references at the bottom of the PDF or in a separate document

Common Diagram Type Templates

Character Relationship Diagram Template

- 家族/组织用 cluster 分组
- 人物用 box/ellipse，重要人物用特殊形状
- 血缘关系用实线，婚姻用红色，其他用虚线
- 添加图例说明符号含义

Concept Map Template

- 顶层概念在上方
- 子概念逐层展开
- is-a 关系用实线，has-a 用虚线
- 使用颜色区分不同类别

Technical Architecture Diagram Template

- 分层架构用 rankdir=TB
- 组件用矩形，服务用圆角矩形
- 依赖关系用箭头
- 关键路径用粗线或特殊颜色

Troubleshooting

Garbled Chinese characters：

graph [fontname="Arial Unicode MS"]
node [fontname="Arial Unicode MS"]
edge [fontname="Arial Unicode MS"]

Diagram is too complex：

• Use concentrate=true to merge identical edges
• Split into multiple subgraphs
• Use different rankdir

Layout is not ideal：

• Adjust ranksep and nodesep
• Use rank=same to force nodes onto the same rank
• Try different splines settings

Version History

• v1.0 (2026-01-02): Initial version
  • Based on the summary of the "Dream of the Red Chamber" character relationship diagram project
  • Supports automatic generation with Graphviz
  • Integrates WebSearch in-depth research
  • TodoWrite task management