/skill-from-repo

Purpose

Generate a Claude Code skill from a GitHub repository using C3.x codebase analysis. Extracts patterns, test examples, how-to guides, configuration patterns, and architectural overview.

Syntax

/skill-from-repo <repo> [--depth <basic|c3x>] [--output <path>] [--three-stream]

Parameters

Parameter	Type	Required	Default	Description
repo	string	Yes	-	GitHub repo (owner/repo or URL)
--depth	string	No	c3x	Analysis depth: basic (1-2 min) or c3x (20-60 min)
--output	string	No	~/.coditect/skills/{name}/	Output directory
--three-stream	flag	No	true	Enable Code + Docs + Insights streams
--include-issues	flag	No	true	Include GitHub issues analysis
--max-issues	integer	No	100	Maximum issues to analyze
--skip	string	No	-	Skip features: patterns,tests,guides,configs,architecture
--languages	string	No	auto	Filter languages (python,javascript,typescript)
--branch	string	No	main	Git branch to analyze

Examples

Basic Usage

# Analyze GitHub repository
/skill-from-repo facebook/react

# With full URL
/skill-from-repo https://github.com/tiangolo/fastapi

Analysis Depth

# Quick analysis (1-2 minutes)
/skill-from-repo django/django --depth basic

# Full C3.x analysis (20-60 minutes)
/skill-from-repo facebook/react --depth c3x

Advanced Usage

# Skip specific features
/skill-from-repo owner/repo --skip patterns,guides

# Filter languages
/skill-from-repo microsoft/TypeScript --languages typescript,javascript

# Custom output
/skill-from-repo owner/repo --output ~/my-skills/myrepo/

# Specific branch
/skill-from-repo owner/repo --branch develop

Three-Stream Analysis

# Full three-stream analysis
/skill-from-repo facebook/react --three-stream

# Code stream only
/skill-from-repo owner/repo --three-stream --skip docs,insights

Output

~/.coditect/skills/{name}/
├── SKILL.md                    # 300+ lines, AI-enhanced
├── patterns/
│   ├── index.md                # Pattern summary
│   ├── singleton.md            # Detected patterns
│   ├── factory.md
│   └── observer.md
├── examples/
│   └── from_tests/             # Extracted test examples
├── guides/
│   ├── getting_started.md      # AI-enhanced guides
│   ├── authentication.md
│   └── deployment.md
├── api_reference/
│   ├── index.md
│   └── modules/                # Per-module docs
├── architecture/
│   ├── overview.md             # Architecture summary
│   ├── dependency_graph.md
│   └── data_flow.md
└── metadata.json

C3.x Analysis Features

Feature	Flag	Description
C3.1 Patterns	--skip patterns	13 GoF patterns, 9 languages
C3.2 Tests	--skip tests	Extract examples from tests
C3.3 Guides	--skip guides	Generate how-to guides
C3.4 Config	--skip configs	Configuration patterns
C3.5 Architecture	--skip architecture	Architecture overview
C3.7 Arch Patterns	included	MVC, microservices, etc.

What Happens

1. Repository Clone: Clone or access repository
2. AST Parsing: Parse code for 9 supported languages
3. C3.1: Detect design patterns
4. C3.2: Extract test examples
5. C3.3: Generate how-to guides
6. C3.4: Extract configuration patterns
7. C3.5: Generate architecture overview
8. C3.7: Detect architectural patterns
9. Three-Stream: Merge Code + Docs + Insights (if enabled)
10. AI Enhancement: Enhance all outputs
11. Quality Check: Validate against CODITECT standards

Success Output

✅ COMMAND COMPLETE: /skill-from-repo

C3.x Analysis Summary:
- [x] Repository: facebook/react
- [x] Analysis depth: c3x
- [x] Languages: JavaScript, TypeScript

C3.x Results:
- [x] C3.1 Patterns: 15 detected (89% confidence)
- [x] C3.2 Tests: 67 examples extracted
- [x] C3.3 Guides: 12 how-to guides
- [x] C3.4 Config: 8 configuration patterns
- [x] C3.5 Architecture: Overview + dependencies
- [x] C3.7 Arch: component-based, event-driven

Three-Stream Analysis:
- [x] Code: AST analysis complete
- [x] Docs: README, CONTRIBUTING extracted
- [x] Insights: 100 issues analyzed, 24 common problems

Output: ~/.coditect/skills/react/
Quality Score: 8.9/10

When NOT to Use

Do NOT use this command when:

• Analyzing documentation websites - use /skill-from-docs instead
• Need to combine docs + code sources - use skill-generator-orchestrator instead
• Repository is private and no GitHub token configured
• Repository requires authentication and no access token available
• Quick one-off code question - just ask Claude directly
• Repository is too large (>100K files) without --languages filter
• Analyzing minified/compiled code (dist/, build/) - garbage patterns
• Need PDF analysis - use /pdf skill instead

Completion Checklist

Before marking this command as complete, verify:

• Repository cloned or local path validated
• C3.1 Design patterns detected and documented
• C3.2 Test examples extracted with quality filtering
• C3.3 How-to guides generated (if depth=c3x)
• C3.4 Configuration patterns extracted
• C3.5 Architecture overview generated
• C3.7 Architectural patterns detected
• SKILL.md generated (300+ lines for c3x, 100+ for basic)
• All output directories created (patterns/, examples/, guides/, etc.)
• metadata.json created with analysis metrics

Failure Indicators

• ❌ Repository not accessible (clone failed, 404, auth required)
• ❌ No supported language files found (9 languages supported)
• ❌ Zero patterns detected (check language filter)
• ❌ SKILL.md under 100 lines
• ❌ GitHub rate limit exceeded (set GITHUB_TOKEN)
• ❌ Zero test examples extracted
• ❌ API reference empty

Anti-Patterns (Avoid)

Anti-Pattern	Problem	Solution
No language filter on huge repos	Memory exhaustion, timeout	Use --languages python,typescript
Skipping test extraction	Missing real-world examples	Keep --skip tests off
Surface-only pattern detection	Low confidence (0.6)	Use --depth c3x for production
Ignoring GitHub insights	Missing community knowledge	Enable --three-stream
No AI enhancement	Basic template output	AI enhancement enabled by default
Analyzing minified code	Garbage patterns	Exclude dist/, build/ via .gitignore
No GITHUB_TOKEN for public repos	Rate limiting (60/hour)	Set env var for 5000/hour

Verification

After execution, verify success:

# 1. Check output structure exists
ls -la ~/.coditect/skills/{name}/

# 2. Verify SKILL.md length
wc -l ~/.coditect/skills/{name}/SKILL.md  # 300+ for c3x, 100+ for basic

# 3. Check pattern detection
ls ~/.coditect/skills/{name}/patterns/

# 4. Validate examples compile (Python example)
python3 -m py_compile ~/.coditect/skills/{name}/examples/from_tests/*.py 2>/dev/null || echo "No Python examples"

# 5. Check architecture output
head -30 ~/.coditect/skills/{name}/architecture/overview.md

# 6. Verify metadata
cat ~/.coditect/skills/{name}/metadata.json | python3 -m json.tool

Troubleshooting

Issue	Solution
Rate limited	Set GITHUB_TOKEN environment variable
Repo too large	Use --languages filter
Missing patterns	Try --depth full for deeper analysis
Timeout	Use --depth basic for quicker analysis

Related

• Agent: codebase-skill-extractor
• Orchestrator: skill-generator-orchestrator
• Companion: /skill-from-docs
• Skill: multi-source-skill-generation