Claude Agentic System

A comprehensive personal agentic system built on top of Claude Code and the Claude Agent SDK, featuring custom skills, slash commands, specialized agents, hooks, output styles, and MCP server integrations.

Overview

This repository provides a complete, production-ready customization framework for Claude Code that enables:

• Skills Library: Autonomous capabilities that Claude invokes based on context
• Slash Commands: User-triggered workflows for common development tasks
• Custom Agents: Specialized AI assistants for specific domains (architecture, security, testing, etc.)
• Hooks: Event-driven automation for tool validation, auditing, and environment setup
• Output Styles: Customizable system prompts for different use cases (research, documentation, teaching, etc.)
• MCP Servers: Pre-configured external tool integrations (Exa search, Cerebras Code, DeepWiki)

Features

Skills

• Collaboration: Brainstorming, planning, pair programming
• Testing: TDD, integration testing, E2E testing, test automation
• Debugging: Systematic debugging, log analysis, performance debugging
• Coding: Clean code, refactoring, design patterns, code review, security
• Architecture: System design, API design, database design, microservices
• Documentation: Technical writing, code documentation, auto-generation
• DevOps: CI/CD, containerization, monitoring
• Security: Security audits, vulnerability scanning, penetration testing
• AI-Assisted: Prompt engineering, AI code generation, research, workflow automation

Commands

• Planning: /brainstorm, /write-plan, /execute-plan, /review-plan
• Development: /feature, /refactor, /debug, /test, /optimize
• Review: /review, /security-review, /performance-review, /pr-review
• Git: /commit, /pr-create, /pr-comments, /changelog
• Documentation: /document, /readme, /api-docs
• Analysis: /analyze-codebase, /analyze-dependencies, /analyze-performance, /analyze-security
• AI Ops: /research, /summarize, /explain
• Utilities: /statusline, /init, /clean, /backup

Agents

• Development: Code Architect, Code Explorer, Code Reviewer, Test Engineer, Debugger, Refactoring Specialist
• Analysis: Security Auditor, Performance Analyst, Dependency Analyzer, Code Quality Analyst
• Documentation: Technical Writer, API Documenter, Tutorial Creator
• Research: Web Researcher, Academic Researcher, Technology Scout
• Creative: Idea Generator, Problem Solver, Innovation Consultant

Hooks

• PreToolUse: Bash validation, file permission checking, security scanning
• PostToolUse: Audit logging, metrics collection, notifications
• SessionStart: Environment setup, project initialization
• SessionEnd: Cleanup, session logging

Output Styles

• Default Variations: Concise, Verbose, Explanatory
• Domain-Specific: Research, Academic, Technical Writing, Creative Writing
• Role-Based: Architect, Security Expert, Code Reviewer, Teacher
• Workflow-Specific: Pair Programming, Rapid Prototyping, Production-Ready

Installation

Option 1: User-Level Installation (All Projects)

Install to ~/.claude/ for use across all projects:

git clone https://github.com/maxmelander/claude-agentic-system.git
cd claude-agentic-system
./install.sh

Option 2: Project-Level Installation (Single Project)

Install to .claude/ for a specific project:

cd your-project
git clone https://github.com/maxmelander/claude-agentic-system.git .claude-system
./scripts/setup/install-project.sh

Option 3: As a Plugin

Install as a Claude Code plugin:

claude plugin install maxmelander/claude-agentic-system

Configuration

Environment Variables

Create a .env file with required API keys:

# Required for Exa MCP server
EXA_API_KEY=your_exa_api_key

# Required for Cerebras Code MCP server
CEREBRAS_API_KEY=your_cerebras_api_key

Settings

Customize behavior in settings.json:

{
  "model": "sonnet",
  "alwaysThinkingEnabled": true,
  "enableAllProjectMcpServers": true,
  "outputStyle": "explanatory",
  "permissions": {
    "allow": ["Bash(npm install:*)", "Bash(find:*)"],
    "deny": ["Read(.env)", "Read(**/.ssh/**)"]
  }
}

Usage

Using Skills

Skills are automatically invoked by Claude based on context. To see available skills:

What skills are available?

Example:

I want to implement a new feature using TDD methodology

Claude will automatically invoke the test-driven-development skill.

Using Commands

Commands are explicitly invoked with / prefix:

/brainstorm I want to build a realtime chat application
/feature Create user authentication with JWT
/review Check this PR for security issues
/commit Create a commit with these changes

Using Agents

Agents can be selected in the agent selector or invoked by Claude:

Can you use the security-auditor agent to review my code?

Using Hooks

Hooks run automatically based on events. View hook configuration in hooks.json.

Using Output Styles

Switch output styles with:

/output-style research
/output-style pair-programming

Directory Structure

claude-agentic-system/
├── .claude-plugin/          # Plugin manifest
├── commands/                # Slash commands
├── agents/                  # Custom subagents
├── skills/                  # Skills library
├── hooks/                   # Hook scripts
├── output-styles/           # Custom output styles
├── mcp-servers/             # MCP configurations
├── templates/               # Configuration templates
├── scripts/                 # Utility scripts
├── docs/                    # Documentation
├── examples/                # Example configurations
└── tests/                   # Tests

Documentation

• Getting Started Guide
• Architecture Overview
• Skills Guide
• Commands Guide
• Agents Guide
• Hooks Guide

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.

To create a new skill:

1. Read docs/skills-guide.md
2. Use the template in templates/skill-template.md
3. Follow the skill creation checklist

To request a new skill:

• Add your request to skills/REQUESTS.md

Testing

Run validation tests:

./scripts/validation/validate-skills.sh
./scripts/validation/validate-commands.sh
./scripts/validation/validate-agents.sh
./scripts/validation/validate-hooks.sh

Maintenance

Backup Configuration

./scripts/maintenance/backup.sh

Update System

./scripts/maintenance/update.sh

Sync Configurations

./scripts/maintenance/sync.sh

Requirements

• Claude Code CLI (claude)
• Git
• Bash (for shell scripts)
• Python 3.8+ (for Python hooks and utilities)
• Node.js (for MCP servers)
• gh CLI (for GitHub operations)

License

MIT License - see LICENSE for details.

Acknowledgments

Built on top of Claude Code and the Claude Agent SDK.

Support

• Report issues: GitHub Issues
• Documentation: docs/
• Examples: examples/

Changelog

See CHANGELOG.md for version history.