autodoc-github-starrepo-skill

name	Autodoc-GitHub-StarRepo-Skill
description	Automatically convert GitHub Star repositories into high-quality English documentation. Input a GitHub repository URL or owner/repo format, automatically fetch DeepWiki documentation and README files, generate refined English documentation, and save to Obsidian. Use cases: "Convert https://github.com/owner/repo to English documentation".

GitHub Repository Documentation Generator

Core Workflow

Fetch Documentation → Generate Draft → Supplement & Refine → Save Output → Discord Notification

Mandatory Pre-Read

Before generating documentation, you MUST review:

• references/doc_template.md - Documentation structure template
• references/deepwiki_mcp.md - DeepWiki API usage (with working code)

Two-Phase Workflow

Phase One: Generate Draft

You MUST read the following files:

1. {WORK_DIR}/overview.md - Overview document generated by DeepWiki
2. {WORK_DIR}/README.md - GitHub README file

Generate files:

• title.txt - English title (10 characters or less)
• draft.md - English draft documentation

Draft requirements:

• Keep technical terms (variable names, function names, class names) in English
• Remove redundant source file reference lists
• Focus on architecture design and core features

Phase Two: Supplement & Refine

Check and supplement:

1. Run ls {WORK_DIR}/docs/ to view all documentation
2. Selectively read relevant documents to supplement gaps in the draft
3. Add missing important details
4. Correct inaccurate technical descriptions

Output:

• final.md - Final documentation

Output Requirements

Documentation structure (see doc_template.md):

• Project introduction (opening paragraph)
• Core features/highlights
• Architecture design
• Technical highlights
• Usage instructions (if applicable)
• Key concept explanations

Classification Framework

GitWiki/
├── AI Agent Frameworks/
│   ├── Multi-Agent/
│   ├── Single-Agent/
│   └── Coding Agents/
├── Deep Research/
│   ├── Agent Research/
│   └── Paper Analysis/
├── Academic Papers/
├── Video & Media/
├── Development/
│   ├── WeChat Mini Programs/
│   ├── Android Development/
│   └── Web Development/
├── Knowledge Management/
├── Learning & Tutorials/
├── Productivity Tools/
├── Desktop Applications/
├── Biomedical/
└── Other/

Output Paths

• Obsidian: {ObsidianVault}/GitWiki/{Category}/{repo}_{title}.md
• Syncthing: {SyncthingDir}/{Category}/{repo}_{title}.md

Discord Channel Notification

✅ Documentation Generated Successfully

Repository: owner/repo
Category: AI Agent Frameworks/Multi-Agent
Title: Project Title
File Path: GitWiki/AI Agent Frameworks/Multi-Agent/owner_repo_ProjectTitle.md

Channel Configuration

Configure Discord channel notifications via environment variables:

• DISCORD_CHANNEL_ID - Discord channel ID (optional, no notification if not set)

Environment Configuration

Variable	Description
GITHUB_TOKEN	GitHub API Token
OBSIDIAN_VAULT_PATH	Obsidian vault path
SYNCTHING_DIR	Syncthing sync directory (optional)
DEEPWIKI_BASE_URL	DeepWiki MCP URL (default: https://mcp.deepwiki.com/mcp)

DeepWiki MCP Usage Example

import json
import urllib.request
import ssl

DEEPWIKI_URL = "https://mcp.deepwiki.com/mcp"

def get_overview(repo_name: str) -> str:
    """Fetch overview document"""
    payload = {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/call",
        "params": {
            "name": "read_wiki_contents",
            "arguments": {"repoName": repo_name}
        }
    }
    data = json.dumps(payload).encode('utf-8')
    context = ssl.create_default_context()
    context.check_hostname = False
    context.verify_mode = ssl.CERT_NONE

    req = urllib.request.Request(DEEPWIKI_URL, data=data, headers={
        "Content-Type": "application/json",
        "Accept": "text/event-stream, application/json"
    })

    with urllib.request.urlopen(req, data=data, timeout=120, context=context) as response:
        content = response.read().decode('utf-8')
        # Parse SSE response, extract result.content[0].text
        return parse_sse_response(content)

Key Tips

1. Generate draft from overview + README first, then review detailed documentation for supplementation
2. MUST use Write tool to create files, not just display in responses
3. MUST generate title.txt, used for final filename
4. Classification based on starred repository analysis, user has 11 category preferences