1. Identify Project Type and Tech Stack
   • Check configuration files:

     package.json

     ,

     pyproject.toml

     ,

     Cargo.toml

     ,

     go.mod

     ,

     pom.xml

     , etc.
   • Identify frameworks: React, Vue, Django, FastAPI, Spring Boot, etc.
   • Identify build tools and dependency managers
2. Analyze Directory Structure
   • Scan main directories using Glob
   • Identify layered patterns: controllers, services, models, utils, etc.
   • Mark entry files and configuration files
3. Extract Core Modules
   • Identify main functional modules
   • Analyze dependency relationships between modules
   • Identify external service dependencies (databases, caches, message queues, etc.)
4. Generate Architecture Diagram

graph TB
    subgraph 表示层["表示层 (Presentation)"]
        UI[用户界面]
        API[API 网关]
    end

    subgraph 业务层["业务层 (Business)"]
        Service1[服务模块1]
        Service2[服务模块2]
    end

    subgraph 数据层["数据层 (Data)"]
        DB[(数据库)]
        Cache[(缓存)]
    end

    UI --> API
    API --> Service1
    API --> Service2
    Service1 --> DB
    Service2 --> Cache

• Layered Architecture: Suitable for traditional MVC, three-tier architecture projects
• Microservices Architecture: Suitable for multi-service, distributed systems
• Frontend-Backend Separation: Suitable for SPA + API projects
• Monolithic Application: Suitable for small-scale projects

1. Define Analysis Objectives
   • Clarify the specific module or function to analyze
   • Identify entry points (API endpoints, event handlers, command entries)
2. Trace Code Call Chains
   • Track function calls starting from the entry point
   • Record data transformations between layers
   • Mark key data processing nodes
3. Identify Data Models
   • Analyze input data structures
   • Trace data transformation processes
   • Record output data structures
4. Generate Sequence Diagrams and Data Flow Diagrams

sequenceDiagram
    participant C as 客户端
    participant A as API层
    participant S as 服务层
    participant D as 数据层

    C->>A: 发送请求
    A->>A: 参数验证
    A->>S: 调用业务逻辑
    S->>D: 查询数据
    D-->>S: 返回数据
    S->>S: 业务处理
    S-->>A: 返回结果
    A-->>C: 响应结果

flowchart LR
    subgraph 输入["输入数据"]
        I1[请求参数]
        I2[用户凭证]
    end

    subgraph 处理["处理过程"]
        P1[验证] --> P2[转换]
        P2 --> P3[业务逻辑]
        P3 --> P4[持久化]
    end

    subgraph 输出["输出数据"]
        O1[响应结果]
        O2[事件通知]
    end

    I1 --> P1
    I2 --> P1
    P4 --> O1
    P3 --> O2

1. Scan the project root directory using Glob to identify configuration files
2. Analyze directory structure to identify main modules
3. Use Grep to search import/require statements for dependency analysis
4. Generate Mermaid architecture diagram after comprehensive analysis
5. Annotate the responsibilities of each module in Chinese
6. Use the

   mermaid-live-preview

   scripting tool to generate online preview links for each Mermaid diagram

1. Confirm the specific module or function the user wants to analyze
2. Locate entry files and entry functions
3. Track function calls layer by layer, using the Read tool to view code
4. Record data changes at each step
5. Generate sequence diagrams and data flow diagrams
6. Describe key data flow nodes in Chinese
7. Use the

   mermaid-live-preview

   scripting tool to generate online preview links for each Mermaid diagram

• Project Name
• Tech Stack
• Project Type

• Main Directory Explanations

• Databases
• Third-party Services

undefined

• Feature/Module Name
• Entry Points

• Input Structure
• Output Structure

---

1. Check docs Directory
   • Create the

     docs/

     directory if it does not exist
   • Check for existing architecture documents to avoid overwriting
2. Determine File Names
   • Architecture analysis:

     docs/architecture.md

   • Data flow analysis:

     docs/dataflow-{module-name}.md

   • Use user-specified file names if provided
3. Write Documents
   • Use the Write tool to save the analysis report to the target file
   • Add generation time and analysis scope notes at the top of the document
4. Generate Online Preview Links
   • For each generated Mermaid diagram, use the

     encode.py

     script from the

     mermaid-live-preview

     skill to generate online preview links:
     bash

     python3 <skills-root>/skills/mermaid-live-preview/scripts/encode.py "<mermaid-code>"

   • Append the output Edit and View links after the corresponding Mermaid code block
   • Format example:
     markdown

     [Edit Online](https://mermaid.live/edit#pako:...) | [View Online](https://mermaid.live/view#pako:...)

5. Confirm Output
   • Inform the user of the saved document path
   • Include online preview links for direct access to the charts

Generation Time: YYYY-MM-DD Analysis Scope: Entire Project

Generation Time: YYYY-MM-DD Analyzed Module: {Module Path}

• Prioritize using existing project documents (README, API docs) as references
• Keep architecture diagrams concise and avoid excessive details
• Focus sequence diagrams on main processes; branch logic can be explained separately
• Use Chinese annotations for all diagrams
• Provide brief responsibility descriptions for each module
• Must include online preview links for each Mermaid diagram (using

  encode.py

  from the

  mermaid-live-preview

  skill)
• Always save analysis results to the

  docs/

  directory after completion
• Retain historical versions or make incremental updates when modifying existing documents