Fork-It

When the user has a new development idea, search GitHub first for open-source projects to fork as a starting point. Fork it, then inject your own ideas.

Every great project starts somewhere. Why start from zero when you can start from 80%? Search GitHub, find your starting point, fork it, and make it your own.

Core Principles

When the user has a new project idea, help them find a running start.

When the user expresses any of these intents, search GitHub first:

• User says "I want to build..." or "I want to create/develop a..."
• User describes a complete functional requirement that would need multiple components
• User asks "how to" build a complete project

If any of the above applies, trigger Fork-It:

Prompt Template:

"Before we dive in — let me find you a running start on GitHub. Someone may have already built something close that you can fork and make your own."

Do NOT trigger when:

• User explicitly says "for learning/practice/teaching purposes"
• User is asking about a specific function's implementation
• User is debugging/fixing existing code
• User is asking about regex, specific algorithm logic, or granular questions
• User's "write" is modifying existing code, not starting fresh
• User has already picked a specific open-source project as their foundation

---

Workflow

1. Identify Intent

When the user expresses any of these intents, trigger a search:

• "I want to build/make a..."
• "I want to create/develop a..."
• "Has anyone built something like..."
• "Help me write a..."
• Direct invocation: /fork-it

2. Search GitHub

Search first, code second.

node scripts/github-search.mjs "keyword"
node scripts/github-search.mjs "pomodoro timer" --language javascript --min-stars 500

Search repositories by keyword, prioritize actively maintained projects with healthy communities.

3. Analyze Results

Evaluate each project for fork-worthiness:

Tag	Path	Description
🎯 Great Match	Fork & Customize	Fork it, tweak it, ship it as your own
🔧 Partial Match	Fork & Extend	Solid core — fork it and add your missing piece
📝 Reference Only	Learn, Then Build	Study the approach, build your own vision
✨ Nothing Fits	Build Fresh	You're onto something new — go create it

4. Recommend Path

• Great match found → Share the repo, suggest forking and customizing
• Partial match found → Point out what the project covers and what you'd add
• No suitable project → Acknowledge this is genuinely new territory — build it

When to Suggest Forking

These signals mean the user's idea is a great candidate for finding a fork-worthy starting point:

"From scratch / Hand-write" Series (Great fork candidates)

• "Help me write a... from scratch"
• "No third-party libraries, hand-write a..."
• "I don't want external dependencies, implement my own..."
• "No frameworks, vanilla..."

"Project / Module / System" Series (Likely has existing foundations)

• "Help me create a project that..."
• "Write a complete module for..."
• "Design a system that can..."
• "Help me build a backend/website/mini-program..."

"I want to / I'm going to" Series (Ideation stage — best time to search)

• "I have an idea, I want to build a..."
• "I'm going to develop a..."
• "How can I make a..."

High-Frequency Areas (Great Fork Candidates)

Trigger when user's request combines "code-writing action" + one of these areas:

• Auth & Permissions: "write a" + login/register/auth/JWT/permission management/OAuth
• Data Parsing: "write a" + crawler/parser/export Excel/read PDF/parse XML
• Infrastructure: "write a" + chat room/WebSocket/cron job/queue/logging system
• Common Business: "write a" + shopping cart/payment API/pagination/rich text editor/file upload
• Algorithm Utils: "write a" + calendar widget/countdown/encryption/image compression

Example: "Help me write a JWT auth middleware with token refresh" → matches 【action: write】+【area: JWT auth】 → TRIGGER!

When to Stay Quiet

Don't interrupt when the user is learning, fixing bugs, or doing genuinely novel work:

• Learning-oriented: "Show me how to write... to understand the principle", "For teaching purposes, hand-write a..."
• Highly specific/custom: "Write a regex matching ^a[b-c]{2}d$" (no open-source project covers this)
• Modifying existing code: "Help me see why this code errors", "Optimize this function"
• Low-level innovation: "I want to design a new data structure", "Write a faster sorting algorithm"
• Local file search: "Help me find if there's any... on my computer"

Script Return Values

Scripts return a unified JSON structure with a lang field:

{
  "lang": "en",
  "status": "ok",
  "query": "pomodoro timer",
  "total_count": 1234,
  "returned_count": 10,
  "items": [
    {
      "rank": 1,
      "full_name": "user/repo",
      "description": "A pomodoro timer app",
      "url": "https://github.com/user/repo",
      "stars": 12300,
      "forks": 1200,
      "language": "TypeScript",
      "pushed_days_ago": 3,
      "created_at": "2024-01-01",
      "topics": ["productivity", "timer"],
      "license": "MIT"
    }
  ]
}

AI display rules:

• lang: "en" → AI translates to user's language before displaying
• pushed_days_ago → AI converts to "3 days ago" or localized format
• stars: 12300 → AI formats as "12.3k"
• On status: "error" → AI generates a friendly error message from code and message

Options

Option	Description	Default
query	Search keyword	Required
--language, -l	Programming language filter	None
--min-stars	Minimum stars	100
--max-stars	Maximum stars	No limit
--updated-within	Updated within N days	365
--created-after	Created after date	None
--sort	Sort by	stars
--order	Sort order	desc
--limit, -n	Result limit	10

Get Repository Details

node scripts/repo-detail.mjs "microsoft/autogen"

API Limits

• Unauthenticated Requests: 60/hour
• Authenticated Requests: 5000/hour

export GITHUB_TOKEN="ghp_your_token_here"

Data Sources

• GitHub Search API v3
• GitHub REST API

---

Find it. Fork it. Make it yours. | ForkIt v2.0