🔌 Skills & Extensions¶

Extensible plugins, skills, and tools for AI coding assistants like Claude Code, Codex, and similar platforms.

📋 Table of Contents¶

Overview
Tools List
Selection Guide
Installation & Usage

Overview¶

Skills and extensions enhance AI coding assistants with: - Memory Systems: Persistent context across sessions - Token Optimization: Reduce costs and improve efficiency - Knowledge Graphs: Structured code understanding - Marketplaces: Discover and install capabilities - Workflow Tools: Specialized development workflows

Tools List¶

Repo	Description	Stars
skydeckai/claude-mem	Auto-capture Claude sessions, compress with AI, inject context
komalvardhanlolugu/caveman	Reduce token usage by 65% through efficient patterns
komalvardhanlolugu/code-review-graph	Knowledge graph reducing tokens 6.8× on reviews, 49× on coding
komalvardhanlolugu/graphify	Multi-modal knowledge graphs for code, docs, papers, images
komalvardhanlolugu/gstack	23 opinionated tools: CEO, Designer, Eng Manager, QA, Release
claudedotcom/claude-code-plugins-plus-skills	Marketplace: 340 plugins, 1367 agent skills, CCPI manager
markNZed/awesome-claude-plugins	Curated collection with security-sweep scanner, CCHub app
komalvardhanlolugu/understand-anything	Multi-agent pipeline building interactive knowledge graphs

Selection Guide¶

By Problem Solved¶

💾 Memory & Context - claude-mem: Automatic session capture and context injection - Best for: Long-term projects, recurring conversations - Token savings: Varies based on session length

💰 Token Optimization - caveman: 65% reduction through communication patterns - code-review-graph: 6.8-49× reduction via knowledge graphs - graphify: Structured understanding reduces repetition - Best for: Large codebases, budget-conscious teams

🗺️ Knowledge Graphs - graphify: Multi-modal (code, docs, papers, images, videos) - code-review-graph: Specialized for code review workflows - understand-anything: Interactive exploration - Best for: Complex projects, onboarding, documentation

🛠️ Comprehensive Toolkits - gstack: 23 tools for complete dev lifecycle - claude-code-plugins-plus-skills: 1367 skills marketplace - awesome-claude-plugins: Curated collection - Best for: Teams wanting ready-made solutions

By Team Size¶

Solo Developer - caveman - Immediate cost savings - claude-mem - Personal context management - graphify - Understand your own codebase

Small Team (2-10) - code-review-graph - Efficient reviews - gstack - Shared workflows - awesome-claude-plugins - Team standards

Enterprise - claude-code-plugins-plus-skills - Marketplace governance - gstack - Role-based tools - graphify - Onboarding and documentation

Installation & Usage¶

Claude Mem - Session Memory¶

# Install
pip install claude-mem

# Configure
export ANTHROPIC_API_KEY=your-key

# Use with Claude Code
claude-mem init

How it works: 1. Capture: Automatically saves Claude conversations 2. Compress: AI summarizes key context and decisions 3. Inject: Adds relevant context to future sessions

Example workflow:

# Day 1: Discuss architecture
claude-code chat "Let's design the user authentication system"

# Day 2: Implementation
claude-mem recall "authentication"
# Injects previous architecture decisions
claude-code chat "Implement the auth system we discussed"

Caveman - Token Reduction¶

# Install as Claude Code skill
cd ~/.claude/skills
git clone https://github.com/komalvardhanlolugu/caveman

# Activate
claude-code skill enable caveman

Communication patterns:

❌ Before (100 tokens):
"Can you please help me refactor this function to make it more efficient
and easier to read? I think there might be some performance issues."

✅ After (35 tokens - 65% reduction):
"Refactor function: optimize performance, improve readability"

Best practices: - Remove pleasantries and filler words - Use imperative mood ("Add feature" not "Can you add feature") - Provide context in structured format - Use code references instead of descriptions

Code Review Graph - Efficient Reviews¶

# Install
npm install -g code-review-graph

# Initialize in project
cd your-project
code-review-graph init

# Generate graph
code-review-graph build

# Use with Claude Code
claude-code --context-graph .code-review-graph/graph.json \
  chat "Review the authentication changes"

Token savings example:

Traditional review (1000 files):
- Read all potentially related files: ~500k tokens
- Cost: $15/review

With knowledge graph:
- Graph identifies 10 relevant files: ~50k tokens
- Cost: $1.50/review
- Savings: 90% or $13.50/review

# Install
pip install graphify-ai

# Process codebase
graphify analyze ./src \
  --include-docs \
  --include-tests \
  --output knowledge-graph.html

# Process research papers
graphify analyze ./papers/*.pdf \
  --type academic \
  --cluster-topics

# Process videos
graphify analyze ./tutorials/*.mp4 \
  --extract-frames \
  --transcribe

# View interactive graph
open knowledge-graph.html

Supported inputs: - Code: Python, JavaScript, TypeScript, Go, Rust, Java, C++, 18 more - Documents: PDF, Markdown, Word, HTML - Images: Architecture diagrams, screenshots, flowcharts - Videos: Tutorials, presentations, demos - Audio: Meeting recordings, podcasts

Output:

{
  "nodes": [
    {"id": "UserService", "type": "class", "file": "src/user.py"},
    {"id": "Database", "type": "module", "file": "src/db.py"}
  ],
  "edges": [
    {"source": "UserService", "target": "Database", "type": "imports"}
  ],
  "clusters": [
    {"name": "Authentication", "nodes": ["UserService", "AuthMiddleware"]}
  ]
}

gstack - Complete Development Toolkit¶

# Install
git clone https://github.com/komalvardhanlolugu/gstack ~/.claude/skills/gstack

# Available tools (23 total)
/office-hours      # CEO: Strategic planning
/plan-ceo-review   # CEO: Review plans
/design-review     # Designer: UI/UX review
/plan-design-review # Designer: Design planning
/plan-eng-review   # Eng Manager: Technical review
/review            # Code review
/ship              # Release management
/canary            # Gradual deployment
/qa                # Quality assurance
/browse            # Web browsing
# ... 13 more

Example workflow:

# 1. CEO: Plan feature
/plan-ceo-review "Add dark mode"

# 2. Design: Review UI
/design-review "Dark mode mockups"

# 3. Engineering: Implement
claude-code chat "Implement dark mode based on plan"

# 4. QA: Test
/qa "Test dark mode across browsers"

# 5. Release: Ship
/ship "Deploy dark mode to production"

# 6. Monitor: Canary
/canary "Monitor dark mode rollout"

Claude Code Plugins Marketplace¶

# Install CCPI (Claude Code Plugin Installer)
npm install -g @claudeai/ccpi

# Browse marketplace
ccpi search "database"

# Install plugin
ccpi install @awesome/postgres-helper

# Install skill
ccpi install skill:code-generator

# List installed
ccpi list

# Update all
ccpi update

Popular plugins (from 340+):

ccpi install @plugins/git-flow      # Git workflow automation
ccpi install @plugins/test-gen      # Auto-generate tests
ccpi install @plugins/docs-builder  # Generate documentation
ccpi install @plugins/security-scan # Security vulnerability scan
ccpi install @plugins/performance   # Performance profiling

Popular skills (from 1367+):

ccpi install skill:api-design       # REST API best practices
ccpi install skill:database-design  # Schema optimization
ccpi install skill:refactor-expert  # Code refactoring patterns
ccpi install skill:debug-master     # Advanced debugging
ccpi install skill:deploy-wizard    # Deployment automation

Awesome Claude Plugins¶

Browse curated collection: https://github.com/markNZed/awesome-claude-plugins

Categories: - 🔒 Security: Vulnerability scanning, secrets detection - ⚡ Performance: Profiling, optimization - 🧪 Testing: Unit, integration, E2E test generation - 📝 Documentation: Auto-docs, API specs - 🎨 Code Quality: Linters, formatters, reviewers

CCHub Desktop App:

# Install CCHub
brew install cchub

# Browse plugins visually
cchub

# Features:
# - Visual plugin browser
# - One-click installation
# - Automatic updates
# - Conflict resolution
# - Usage analytics

Token Optimization Comparison¶

Real-World Benchmarks¶

Scenario: Code Review (1000-file project)

Method	Tokens Used	Cost (@$15/M)	Time
No optimization	500,000	$7.50	5 min
caveman patterns	175,000	$2.63	5 min
code-review-graph	50,000	$0.75	3 min
caveman + graph	17,500	$0.26	3 min

Savings: $7.24/review (97%) with combined approach

Scenario: Daily Coding Session (8 hours)

Method	Tokens Used	Cost	Context Quality
No memory	2,000,000	$30	Low (repeats)
claude-mem	400,000	$6	High (recalls)
graphify	300,000	$4.50	Very High

Savings: $25.50/day (85%) with claude-mem + graphify

Architecture Examples¶

1. Memory-Optimized Setup¶

claude-mem (Session capture)
    ↓
caveman (Compress prompts)
    ↓
Claude Code (Execute)
    ↓
claude-mem (Store results)

Best for: Long-term projects, consultants

2. Knowledge Graph Workflow¶

graphify (Build graph)
    ↓
code-review-graph (Optimize reviews)
    ↓
Claude Code (Focused analysis)
    ↓
graphify (Update graph)

Best for: Large codebases, teams

3. Full Stack Development¶

gstack /plan-ceo-review (Strategy)
    ↓
gstack /design-review (Design)
    ↓
Claude Code + graphify (Implementation)
    ↓
gstack /qa (Testing)
    ↓
gstack /ship (Deployment)

Best for: Complete feature development

Best Practices¶

1. Start Simple, Add Complexity¶

Week 1: Install caveman (immediate savings)
Week 2: Add claude-mem (context benefits)
Week 3: Integrate graphify (full optimization)

2. Measure Impact¶

# Before
claude-code stats --tokens

# After caveman
claude-code stats --tokens
# Calculate savings

3. Combine Tools¶

caveman + claude-mem: Efficient memory
graphify + code-review-graph: Complete understanding
gstack + marketplace: No redundant tools

4. Update Regularly¶

# Weekly
ccpi update
git -C ~/.claude/skills/gstack pull

# Monthly
graphify rebuild  # Refresh knowledge graphs

Common Use Cases¶

Onboarding New Developers¶

Stack: graphify + understand-anything 1. Generate interactive codebase graph 2. New dev explores visually 3. Ask questions with full context 4. Faster ramp-up (days vs weeks)

Cost-Conscious Startup¶

Stack: caveman + code-review-graph + claude-mem - Reduce API costs by 85-97% - Maintain quality - Scale usage without scaling costs

Enterprise Compliance¶

Stack: gstack + marketplace governance - Approved plugins only - Audit trail - Role-based access - Security scanning

Research & Learning¶

Stack: graphify + understand-anything - Process papers, videos, code - Build knowledge graphs - Interactive exploration - Multi-modal learning

Security Considerations¶

Local vs Cloud¶

Local Tools (graphify, code-review-graph): - ✅ Code stays on your machine - ✅ No external API calls for graph building - ❌ Graph visualization may use CDNs

Cloud Tools (claude-mem with API): - ⚠️ Sessions sent to compression service - ✅ Can self-host compression model - ⚠️ Check data retention policies

Marketplace (claude-code-plugins-plus-skills): - ⚠️ Review plugin permissions - ✅ Community security scanning - ⚠️ Update promptly for security patches

Best Practices¶

Review Plugin Code

ccpi inspect @plugin/name
# Check file access, network calls, credentials

Use Security Scanner

# From awesome-claude-plugins
security-sweep scan ~/.claude/skills

Limit Permissions

# ~/.claude/config.yml
plugins:
  allow_file_write: false
  allow_network: false
  allowed_domains: ["github.com", "anthropic.com"]

Developer Tools - Base AI coding assistants
MCP Servers - Model Context Protocol integrations
Observability - Monitor skill usage
Agents & Orchestration - Build on skills

← Back to Home | Next: MCP Servers →

🔌 Skills & Extensions¶

📋 Table of Contents¶

Overview¶

Tools List¶

Selection Guide¶

By Problem Solved¶

By Team Size¶

Installation & Usage¶

Claude Mem - Session Memory¶

Caveman - Token Reduction¶

Code Review Graph - Efficient Reviews¶

Graphify - Multi-Modal Knowledge Graphs¶

gstack - Complete Development Toolkit¶

Claude Code Plugins Marketplace¶

Awesome Claude Plugins¶

Token Optimization Comparison¶

Real-World Benchmarks¶

Architecture Examples¶

1. Memory-Optimized Setup¶

2. Knowledge Graph Workflow¶

3. Full Stack Development¶

Best Practices¶

1. Start Simple, Add Complexity¶

2. Measure Impact¶

3. Combine Tools¶

4. Update Regularly¶

Common Use Cases¶

Onboarding New Developers¶

Cost-Conscious Startup¶

Enterprise Compliance¶

Research & Learning¶

Security Considerations¶

Local vs Cloud¶

Best Practices¶

Related Resources¶