Master Claude Code in a Week (Part 10): Plugins & Advanced Features
·
5 min read
This is the final part of the "Master Claude Code in a Week" series. In this article, we'll explore Plugins and Advanced Features of Claude Code.
What are Plugins?
Plugins are bundled packages of all features we've learned: CLAUDE.md, Memory, Hooks, MCP, and Subagents combined into shareable packages for teams or communities.
┌────────────────────────────────────────────────────────────────┐
│ PLUGIN ARCHITECTURE │
├────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────── Plugin Package ────────────┐ │
│ │ │ │
│ │ ┌─────────────────────────────────┐ │ │
│ │ │ CLAUDE.md │ │ Instructions │
│ │ │ (Project Instructions) │ │ │
│ │ └─────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────┐ │ │
│ │ │ Memory/ │ │ Context │
│ │ │ (Project Memories) │ │ │
│ │ └─────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────┐ │ │
│ │ │ Hooks/ │ │ Automation │
│ │ │ (Event Handlers) │ │ │
│ │ └─────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────┐ │ │
│ │ │ Agents/ │ │ Specialization │
│ │ │ (Subagents) │ │ │
│ │ └─────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────┐ │ │
│ │ │ settings.json │ │ MCP Config │
│ │ │ (MCP Servers) │ │ │
│ │ └─────────────────────────────────┘ │ │
│ │ │ │
│ └────────────────────────────────────────┘ │
│ │
└────────────────────────────────────────────────────────────────┘
Creating a Plugin
Structure
my-plugin/
├── README.md # Documentation
├── install.sh # Installation script
├── CLAUDE.md # Instructions template
├── memory/
│ ├── patterns.md
│ └── decisions.md
├── hooks/
│ ├── pre-commit.sh
│ └── post-edit.py
├── agents/
│ ├── code-reviewer.md
│ └── test-engineer.md
└── mcp/
└── settings.json # MCP configuration
Example: Laravel Plugin
install.sh:
#!/bin/bash
# Laravel Plugin Installation
echo "Installing Laravel Plugin for Claude Code..."
# Create directories
mkdir -p .claude/{memory,hooks,agents}
# Copy CLAUDE.md
cp templates/CLAUDE.md ./CLAUDE.md
# Copy memory files
cp memory/*.md .claude/memory/
# Copy hooks
cp hooks/*.sh .claude/hooks/
chmod +x .claude/hooks/*.sh
# Copy agents
cp agents/*.md .claude/agents/
# Add MCP configuration
if [ -f ~/.claude/settings.json ]; then
echo "Merging MCP configuration..."
# Merge logic here
fi
echo "Done! Plugin installed."
Installation
# Clone plugin repository
git clone https://github.com/example/laravel-claude-plugin
# Run installation
cd laravel-claude-plugin
./install.sh
Advanced Features
1. Planning Mode
Planning Mode is for complex tasks that need step-by-step planning.
Enable with:
/plan enable
Or use the CLI:
claude -p "Refactor this authentication system"
Workflow:
┌─────────────────────────────────────────────────────────────────┐
│ PLANNING MODE │
├─────────────────────────────────────────────────────────────────┤
│ │
│ 1. Analyze → Claude analyzes the task's requirements │
│ │
│ 2. Plan → Claude creates a detailed execution plan │
│ │
│ 3. Review → You review and approve/modify the plan │
│ │
│ 4. Execute → Claude executes step by step │
│ │
│ 5. Report → Summary of completed work │
│ │
└─────────────────────────────────────────────────────────────────┘
Use Cases:
- Major refactoring
- New feature implementation
- Complex debugging
- System architecture changes
2. Extended Thinking
Extended Thinking gives Claude more "thinking time" for complex problems.
Enable:
/config set thinking_mode extended
Or add to CLAUDE.md:
## Claude Settings
- Use extended thinking for complex analysis
- Don't rush to conclusions
When to Use:
- Complex debugging
- Performance analysis
- Architecture design
3. Background Tasks
Background Tasks let Claude continue working without blocking the terminal.
# Start a long task
/background "Analyze all API endpoints and document them"
# View progress
/background status
# List running tasks
/background list
Ideal for:
- Documentation generation
- Code analysis
- Large refactoring
4. Permission Modes
Claude Code operates in different permission modes:
┌───────────────────────────────────────────────────────────────┐
│ PERMISSION MODES │
├───────────────────────────────────────────────────────────────┤
│ │
│ Supervised Mode (Default) │
│ ├── Ask before major file changes │
│ ├── Confirm before running commands │
│ └── Safe for normal development │
│ │
│ Auto-Edit Mode │
│ ├── Automatic file edits within project │
│ ├── Still asks for command execution │
│ └── Good for well-defined tasks │
│ │
│ YOLO Mode (Full Auto) │
│ ├── No confirmations needed │
│ ├── Claude executes freely │
│ └── Use with caution! Only for trusted tasks │
│ │
└───────────────────────────────────────────────────────────────┘
Configure:
# Terminal
/permissions supervised # Default
/permissions auto-edit # Auto-edit files
/permissions yolo # Full automatic
# Or in settings
claude config set --global permission_mode auto-edit
5. Session Management
View Session:
/session info
Export Session:
# Export for sharing later
/session export my-session.json
Resume Session:
# Resume from file
claude --resume my-session.json
6. Headless Mode
Run Claude without interactive terminal:
# Execute single task
claude -p "Update all tests to match new API" --output-format json
# Read from stdin
cat task.md | claude --stdin
Useful for:
- CI/CD pipelines
- Batch processing
- Automation scripts
7. Git Integration
Claude Code has deep Git integration:
# Review current changes
/git diff
# Create commit with auto-generated message
/git commit
# Analyze PR
/git pr-review
Auto-commit on task completion:
Add to CLAUDE.md:
## Git Workflow
After completion:
1. Run tests
2. If all pass, create commit
3. Use conventional commit message format
Combining All Features
Example workflow combining all learned features:
┌────────────────────────────────────────────────────────────────┐
│ COMPLETE DEVELOPMENT WORKFLOW │
├────────────────────────────────────────────────────────────────┤
│ │
│ 1. PLANNING │
│ └── Enable Planning Mode for complex task │
│ │
│ 2. CONTEXT │
│ ├── Load CLAUDE.md (project rules) │
│ └── Load Memory (past decisions) │
│ │
│ 3. DATA │
│ ├── MCP GitHub: fetch issues/PRs │
│ └── MCP Database: query schemas │
│ │
│ 4. IMPLEMENT │
│ ├── Main agent coordinates │
│ ├── Delegate to @test-engineer │
│ └── Delegate to @code-reviewer │
│ │
│ 5. QUALITY │
│ ├── Hook: pre-commit checks │
│ ├── Run tests automatically │
│ └── Security scan │
│ │
│ 6. CHECKPOINT │
│ └── Create checkpoint before merge │
│ │
│ 7. COMPLETE │
│ ├── Save to memory │
│ └── Auto-commit changes │
│ │
└────────────────────────────────────────────────────────────────┘
Summary: Complete Learning Path
We've covered Claude Code's complete feature set:
| Part | Topic | What You Learned |
|---|---|---|
| 1 | Introduction | Overview & installation |
| 2 | Slash Commands | Quick actions |
| 3 | Memory | Context persistence |
| 4 | Checkpoints | Save/restore work |
| 5 | CLI | Command-line power |
| 6 | Skills | Reusable templates |
| 7 | Hooks | Event automation |
| 8 | MCP | External tool integration |
| 9 | Subagents | Specialized assistants |
| 10 | Plugins | Feature bundles |
Recommended Practice Path
Week 1: Days 1-2
└── Basic usage, Slash Commands
Week 1: Days 3-4
└── Memory, Checkpoints, CLI
Week 1: Days 5-7
└── Skills, Hooks, MCP
Week 2:
└── Subagents, Plugins, Advanced workflows
Ongoing:
└── Customize for your team
└── Create and share plugins
└── Contribute to claude-howto
Resources
- Official Docs: claude.ai/code
- claude-howto Repository: github.com/luongnv89/claude-howto
- Community Plugins: awesome-claude-code
Conclusion
Claude Code is a powerful tool that combines AI capabilities with practical developer workflows. The key is:
- Start simple - Use basic commands first
- Add gradually - Introduce Memory, then Hooks, etc.
- Customize - Adapt to your project's needs
- Share - Create plugins for your team
Thank you for following this series! Happy coding with Claude! 🚀
This series is translated and expanded from claude-howto — MIT License.