name	study-from-github
description	Learn undocumented or poorly-documented SDKs/APIs by studying real-world code on GitHub. Use when asked to "learn from GitHub", "study source code", "find examples", "understand unknown API", "check how others use X", "look for patterns".

Study from GitHub

Learn how to use unknown or poorly-documented APIs by searching, cloning, and studying real-world code on GitHub.

When to Use This Skill

• Apple's/Xcode/framework API lacks documentation or examples
• Need working code patterns for an SDK
• "How do I use X?" when search engines don't give clear answers
• Learning a new library by example

The Process

Step 1: Search GitHub for Real Code

Use gh search code to find repositories using the API:

gh search code "<API-NAME>" --language swift --limit 10

Pro tip: Combine API name with related terms to narrow results:

gh search code "RecognizeDocumentsRequest table" --language swift --limit 10
gh search code "VNRecognizeTextRequest boundingBox" --language swift --limit 10
gh search code "PDFPage thumbnail OCR" --language swift --limit 10

Examples:

• RecognizeDocumentsRequest
• VNRecognizeTextRequest
• PDFPage.thumbnail

Tips:

• Sort by repos with most stars (likely well-maintained)
• Filter by language matching your project (swift, python, etc.)
• Add --limit 20 if results are sparse

Step 2: Evaluate Search Results

For each promising result, check:

• Repository purpose — Is it related to your use case?
• Star count — Higher stars = more likely battle-tested code
• File path — Usually Sources/, Lib/, or root level

Prioritize repos that:

• Are apps/projects (not forks or archived)
• Have recent commits
• Match your domain (e.g., document scanning → receipt/OCR apps)

Step 3: Clone and Study

git clone https://github.com/<owner>/<repo>.git

Then find the relevant file:

grep -n "API-NAME" <repo>/**/*.swift

Step 4: Extract Patterns

When studying the code, focus on:

1. How is it initialized? (construction patterns)
2. What configuration options exist? (properties, enums)
3. How is it used in a complete flow? (from input to output)
4. Error handling (what can go wrong?)
5. Platform/version requirements (macOS 12+, iOS 16+, etc.)

Step 5: Test and Adapt

• Copy the relevant code pattern
• Adapt to your project structure
• Test with your own data/files
• Iterate based on results

Example Workflow

Goal: Learn how to extract tables from images using Apple's Vision framework.

Step 1: Search

gh search code "RecognizeDocumentsRequest" --language swift --limit 10

Step 2: Find promising results

• cameronrye/clarissa — Document OCR service
• LuegM/ReceiptBro — Receipt processing with table detection

Step 3: Clone the best match

git clone https://github.com/LuegM/ReceiptBro.git

Step 4: Study the relevant code

cat ReceiptBro/ReceiptBro/Services/OCRService.swift

Step 5: Extract the pattern

let request = RecognizeDocumentsRequest()
let observations = try await request.perform(on: imageData)

if let document = observations.first?.document {
    let tables = document.tables
    // ... process tables
}

Rules

1. No exa/web search for code — Use gh search code only
2. Clone with git, not gh — gh repo clone may fail due to SSH keys
3. Focus on patterns, not theory — Find working code, not documentation
4. Validate before sharing — Test the pattern before recommending it
5. Note version requirements — Many Vision APIs require macOS 26+

Version Requirements

Always check platform requirements:

• RecognizeDocumentsRequest → macOS 26+, iOS 26+
• VNRecognizeTextRequest → macOS 10.15+, iOS 13+
• VNDetectDocumentSegmentationRequest → macOS 13+, iOS 16+

If API requires newer OS than target, suggest fallback or alternative approach.