• Quick Start
• The Guide
• Repository Contents
• Installation
• Contributing
• Sources

🚀 Quick Start

# Clone this repo
git clone https://github.com/TheDecipherist/claude-code-mastery.git
cd claude-code-mastery

# Copy hooks to your Claude config
mkdir -p ~/.claude/hooks
cp hooks/* ~/.claude/hooks/
chmod +x ~/.claude/hooks/*.sh

# Copy the settings template (review and customize first!)
cp templates/settings.json ~/.claude/settings.json

# Copy skills
mkdir -p ~/.claude/skills
cp -r skills/* ~/.claude/skills/

📖 The Guide

📱 Read V3 on GitHub Pages | 📄 View GUIDE.md (V3)

What's Covered

| Part | Topic | Key Takeaway | |------|-------|--------------| | 1 | Global CLAUDE.md as Security Gatekeeper | Define once, inherit everywhere | | 2 | Project Scaffolding Rules | Every project follows same structure | | 3 | MCP Servers | External tool integrations | | 4 | Context7 | Live documentation access | | 5 | Custom Commands | Workflow automation | | 6 | Single-Purpose Chats | 39% degradation from topic mixing | | 7 | Skills & Hooks | Enforcement over suggestion |

📁 Repository Contents

claude-code-mastery/
├── GUIDE.md                    # The complete guide
├── templates/
│   ├── global-claude.md        # ~/.claude/CLAUDE.md template
│   ├── project-claude.md       # ./CLAUDE.md starter
│   ├── settings.json           # Hook configuration template
│   └── .gitignore              # Recommended .gitignore
├── hooks/
│   ├── block-secrets.py        # PreToolUse: Block .env access
│   ├── block-dangerous-commands.sh  # PreToolUse: Block rm -rf, etc.
│   ├── end-of-turn.sh          # Stop: Quality gates
│   ├── after-edit.sh           # PostToolUse: Run formatters
│   └── notify.sh               # Notification: Desktop alerts
├── skills/
│   ├── commit-messages/        # Generate conventional commits
│   │   └── SKILL.md
│   └── security-audit/         # Security vulnerability checks
│       └── SKILL.md
└── commands/
    ├── new-project.md          # /new-project scaffold
    ├── security-check.md       # /security-check audit
    └── pre-commit.md           # /pre-commit quality gates

🔧 Installation

Prerequisites

• Claude Code installed
• Python 3.8+ (for Python hooks)
• jq (for JSON parsing in shell hooks)

Step-by-Step

1. Install Hooks

# Create hooks directory
mkdir -p ~/.claude/hooks

# Copy hook scripts
cp hooks/block-secrets.py ~/.claude/hooks/
cp hooks/block-dangerous-commands.sh ~/.claude/hooks/
cp hooks/end-of-turn.sh ~/.claude/hooks/

# Make shell scripts executable
chmod +x ~/.claude/hooks/*.sh

2. Configure Settings

# If you don't have settings.json yet
cp templates/settings.json ~/.claude/settings.json

# If you already have settings.json, merge the hooks section manually

3. Install Skills

# Create skills directory
mkdir -p ~/.claude/skills

# Copy skills
cp -r skills/* ~/.claude/skills/

4. Set Up Global CLAUDE.md

# Copy template
cp templates/global-claude.md ~/.claude/CLAUDE.md

# Customize with your details
$EDITOR ~/.claude/CLAUDE.md

5. Verify Installation

# Start Claude Code
claude

# Check hooks are loaded
/hooks

# Check skills are loaded
/skills

🔒 Why Hooks Matter

CLAUDE.md rules are suggestions. Hooks are enforcement.

CLAUDE.md saying "don't edit .env"
  → Parsed by LLM
  → Weighed against other context
  → Maybe followed

PreToolUse hook blocking .env edits
  → Always runs
  → Returns exit code 2
  → Operation blocked. Period.

Real-world example from a community member:

"My PreToolUse hook blocks Claude from accessing secrets (.env files) a few times per week. Claude does not respect CLAUDE.md rules very rigorously."

Hook Exit Codes

| Code | Meaning | |------|---------| | 0 | Success, allow operation | | 1 | Error (shown to user only) | | 2 | Block operation, feed stderr to Claude |

🧠 Why Single-Purpose Chats

Research consistently shows topic mixing destroys accuracy:

| Study | Finding | |-------|---------| | Multi-turn conversations | 39% performance drop when mixing topics | | Context rot | Recall decreases as context grows | | Context pollution | 2% early misalignment → 40% failure rate |

Golden Rule: One Task, One Chat

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Add your hooks, skills, or improvements
4. Submit a PR with description

Ideas for Contributions

• [ ] More language-specific hooks (Go, Rust, Ruby)
• [ ] Additional skills (code review, documentation, testing)
• [ ] Framework-specific scaffolding templates
• [ ] MCP server configuration examples

📚 Sources

Official Documentation

• Claude Code Best Practices — Anthropic
• Effective Context Engineering — Anthropic
• Agent Skills — Claude Code Docs
• Hooks Reference — Claude Code Docs

Research

• LLMs Get Lost In Multi-Turn Conversation — arXiv
• Context Rot Research — Chroma
• Claude Loads Secrets Without Permission — Knostic

Community

• Claude Code Hooks: Guardrails That Actually Work
• Claude Code Hooks Mastery
• Claude Code Security Best Practices

📄 License

MIT License - See LICENSE

Built with ❤️ by TheDecipherist and the Claude Code community