deepwiki

📁 huguangju/skills 📅 12 days ago
4
总安装量
4
周安装量
#49773
全站排名
安装命令
npx skills add https://github.com/huguangju/skills --skill deepwiki

Agent 安装分布

opencode 4
gemini-cli 3
claude-code 3
github-copilot 3
kimi-cli 3

Skill 文档

DeepWiki Documentation Generator

Generate DeepWiki-style documentation for local codebases with a clear, repeatable pipeline and explicit outputs.


When to Use

Use this Skill when you need to:

  • Produce structured codebase documentation for onboarding or long-term maintenance
  • Capture architecture, modules, and key flows using consistent templates

Do NOT use this Skill when:

  • You only need API reference docs (use TypeDoc/JSDoc)
  • The codebase is trivial (< 100 lines)
  • A proprietary tool format is required

Core Principles

  1. Big picture before details
  2. Responsibilities before implementation
  3. Semantic-level explanation (avoid line‑by‑line translation)
  4. Use domain terminology, not tutorial tone

Output Contract

docs/
├── overview.md              # Project overview: what it is, tech stack, entry points
├── architecture.md          # Architecture design: module division, dependencies, diagrams
├── modules/                 # Module-level documentation
│   ├── [module-name].md     # Independent documentation for each core module
│   └── ...
├── flows.md                 # Behavior and flows: how the system runs
├── design-decisions.md      # Design decisions: why designed this way
└── appendix.md              # Glossary and references (optional)

Each page must be independently readable and RAG‑friendly.


Workflow (Five-Phase Analysis)

Phase 1: Codebase Reconnaissance

Goal: Identify project type and tech stack.

Actions:

  • Scan directory structure
  • Identify primary language/framework/build system
  • Determine project type (app / library / tool / mono‑repo)

Ignored by default:

  • node_modules/, dist/, build/, .git/
  • Generated/cache directories

Output:

  • Overview draft + tech stack tags

Phase 2: Structure Modeling

Goal: Explain how the project is organized.

Actions:

  • System/subsystem decomposition
  • Module boundaries from directory + dependencies
  • Identify core vs peripheral modules

Output:

  • architecture.md
  • Dependency diagrams (Mermaid)
  • System/subsystem hierarchy table

Template: use templates.md.


Phase 3: Module Understanding

Goal: Explain each core module’s purpose and boundaries.

Actions:

  • Extract key entities (classes/functions/interfaces)
  • Summarize responsibilities and public API
  • Identify common usage patterns

Output:

  • modules/[module-name].md
  • Key entity relationship diagrams

Template: use templates.md.


Phase 4: Flow Synthesis

Goal: Describe how the system runs end‑to‑end.

Actions:

  • Find entry points (CLI / main / server / job)
  • Trace main call chains
  • Summarize critical flows and exceptions

Output:

  • flows.md

Template: use templates.md.


Phase 5: Design Insight

Goal: Explain why the system is designed this way.

Actions:

  • Summarize trade‑offs and design patterns
  • Mark inferred reasoning vs confirmed facts

Output:

  • design-decisions.md

Template: use templates.md.


Quality Gates

  • Every page answers: purpose / boundary / when to care
  • No line‑by‑line code narration
  • Mermaid diagrams render correctly
  • Internal links are valid
  • Terminology is consistent

References

Topic Description Reference
Templates Page templates for each document reference
Mermaid Guide Diagram conventions reference
VitePress Config Docs site setup reference
Framework Patterns Framework-specific analysis focus reference