Claude Code Config

Pre-configured Claude Code setup with specialized agents and workflow automation.

Requirements

• Claude Code v2.1.19 or higher - This configuration uses v2.1.19 features (see below)
• uv - Fast Python package manager (used for running hook scripts)

Claude Code v2.1.19 Features Used

This configuration leverages these features:

• Agent-scoped hooks - Hooks defined in agent frontmatter (e.g., PreToolUse in git-expert) (v2.1.0)
• allowed-tools - Tool restrictions in agent frontmatter (e.g., docs-fetcher is read-only) (v2.1.0)
• additionalContext in PreToolUse - Provides guidance when blocking commands (v2.1.9)
• Task management system - Built-in task tracking with TaskCreate, TaskUpdate, TaskList, TaskGet for complex workflows (v2.1.16)
• Slash command argument syntax - Bracket syntax $ARGUMENTS[0] and shorthand $0, $1 for positional arguments (v2.1.19)

Note: context: fork was evaluated but not used due to compatibility issues with multi-phase workflows.

Quick Start (Plugins)

The easiest way to use this repository's features is via plugins:

1. Add the marketplace

/plugin marketplace add myk-org/claude-code-config

2. Install plugins

# GitHub operations (PR reviews, releases, review handling)
/plugin install myk-github@myk-org

# Local code review and database queries
/plugin install myk-review@myk-org

# Qodo AI code review integration
/plugin install myk-qodo@myk-org

# Cursor agent CLI integration
/plugin install myk-cursor@myk-org

3. Install CLI (for github and review plugins)

uv tool install myk-claude-tools

Or from this repository:

uv tool install git+https://github.com/myk-org/claude-code-config

Available Plugin Commands

Plugin	Command	Description
myk-github	/myk-github:pr-review	Review PR and post inline comments
myk-github	/myk-github:release	Create release with changelog
myk-github	/myk-github:review-handler	Process all review sources
myk-review	/myk-review:local	Review uncommitted changes
myk-review	/myk-review:query-db	Query review database
myk-qodo	/myk-qodo:review	AI-powered code review
myk-qodo	/myk-qodo:describe	Generate PR descriptions
myk-qodo	/myk-qodo:improve	Get improvement suggestions
myk-qodo	/myk-qodo:ask	Ask questions about code
myk-cursor	/myk-cursor:prompt	Run one-shot prompts via Cursor agent CLI

For full orchestrator pattern with agents and hooks, see Full Installation below.

Installation

Clone to any location and symlink into ~/.claude:

# Clone to ~/git/ (or your preferred location)
git clone https://github.com/myk-org/claude-code-config ~/git/claude-code-config

# Create ~/.claude if it doesn't exist
mkdir -p ~/.claude

# Symlink each component
ln -sf ~/git/claude-code-config/agents ~/.claude/agents
ln -sf ~/git/claude-code-config/rules ~/.claude/rules
ln -sf ~/git/claude-code-config/scripts ~/.claude/scripts
ln -sf ~/git/claude-code-config/skills ~/.claude/skills
ln -sf ~/git/claude-code-config/settings.json ~/.claude/settings.json
ln -sf ~/git/claude-code-config/statusline.sh ~/.claude/statusline.sh

Updating:

cd ~/git/claude-code-config && git pull

Your symlinks will automatically point to the updated files.

What's Included

• 4 plugins with 10 commands (myk-github, myk-review, myk-qodo, myk-cursor)
• CLI tool (myk-claude-tools) for plugin operations
• 16 specialized agents for different domains (Python, Go, Java, Docker, Kubernetes, Git, etc.)
• 1 skill for context-aware automation
• Orchestrator pattern with automatic agent routing via CLAUDE.md
• Pre-commit hooks for rule enforcement
• Status line integration
• SessionStart tool validation - Checks for required tools (uv, gh, prek, mcpl) and prompts to install missing ones

Agents

Agent	Purpose
python-expert	Python development, testing, async patterns
go-expert	Go development, goroutines, modules
java-expert	Java/Spring Boot development
frontend-expert	JS/TS/React/Vue/Angular
bash-expert	Shell scripting and automation
docker-expert	Dockerfile, container orchestration
kubernetes-expert	K8s/OpenShift, Helm, GitOps
jenkins-expert	CI/CD pipelines, Jenkinsfiles
git-expert	Git operations, branching strategies
github-expert	GitHub platform operations (PRs, issues, releases)
test-automator	Test suites, CI pipelines
test-runner	Test execution and reporting
debugger	Error analysis, debugging
technical-documentation-writer	Documentation
api-documenter	OpenAPI/Swagger specs
docs-fetcher	Fetches external library/framework documentation, prioritizes llms.txt

Marketplace Plugins

This configuration uses plugins from the Anthropic official marketplace.

Critical (required for the code-review loop): pr-review-toolkit, superpowers, feature-dev All others are optional enhancements.

Install all at once

/plugin marketplace add claude-plugins-official

Then install each plugin:

# --- Critical (required for code-review loop) ---
/plugin install feature-dev@claude-plugins-official
/plugin install pr-review-toolkit@claude-plugins-official
/plugin install superpowers@claude-plugins-official

# --- Optional enhancements ---
/plugin install claude-code-setup@claude-plugins-official
/plugin install claude-md-management@claude-plugins-official
/plugin install code-review@claude-plugins-official
/plugin install code-simplifier@claude-plugins-official
/plugin install coderabbit@claude-plugins-official
/plugin install commit-commands@claude-plugins-official
/plugin install frontend-design@claude-plugins-official
/plugin install github@claude-plugins-official
/plugin install gopls-lsp@claude-plugins-official
/plugin install jdtls-lsp@claude-plugins-official
/plugin install lua-lsp@claude-plugins-official
/plugin install playground@claude-plugins-official
/plugin install pyright-lsp@claude-plugins-official
/plugin install security-guidance@claude-plugins-official

Plugin purpose

Plugin	Purpose
claude-code-setup	Claude Code automation recommendations
claude-md-management	CLAUDE.md file management and improvement
code-review	Pull request code review
code-simplifier	Code simplification and refactoring
coderabbit	CodeRabbit AI code review integration
commit-commands	Git commit, push, PR, and branch cleanup
feature-dev	Feature architecture, codebase exploration, code review
frontend-design	Frontend interface design
github	GitHub operations and workflows
gopls-lsp	Go language server (gopls) integration
jdtls-lsp	Java language server (Eclipse JDT.LS) integration
lua-lsp	Lua language server integration
playground	Interactive HTML playground creation
pr-review-toolkit	PR review with specialized agents (code-reviewer, test-analyzer, silent-failure-hunter)
pyright-lsp	Python language server (Pyright) integration
security-guidance	Security best practices
superpowers	Brainstorming, debugging, TDD, code review workflows

Note: Missing plugins are detected automatically at session start. Claude Code will prompt you to install any missing plugins.

Automatic Documentation Fetching

When specialist agents work with external libraries or frameworks, they can automatically fetch the latest documentation through the docs-fetcher agent. This ensures that code follows current best practices and uses up-to-date APIs.

Key features:

• Prioritizes llms.txt files (LLM-optimized documentation)
• Falls back to official documentation sites
• Caches results for faster subsequent access
• Provides context-relevant excerpts to specialist agents

Example workflow:

python-expert working with FastAPI
         ↓
    docs-fetcher fetches FastAPI docs
         ↓
python-expert uses current best practices

Skills

Skills are similar to slash commands but auto-invoke based on task context rather than requiring explicit invocation.

Skill	Description
agent-browser	Browser automation for web testing, form filling, screenshots, and data extraction

agent-browser Installation

See agent-browser for installation instructions.

MCP Server Access

This configuration uses mcp-launchpad (mcpl) for on-demand MCP server access.

Benefits over native MCP loading:

• Tools are NOT loaded into context at session start
• No 30% context consumption from tool definitions
• Agents discover and call tools on-demand via CLI

Installation:

uv tool install https://github.com/kenneth-liao/mcp-launchpad.git

See rules/15-mcp-launchpad.md for detailed usage instructions and command reference.

Why Agent-Based Workflow?

This configuration implements an orchestrator pattern where Claude acts as a manager delegating to specialist agents. Here's why:

Context Preservation

• Main conversation stays lean - The orchestrator only tracks high-level progress
• Specialists work in isolation - Each agent handles its task without bloating the main context
• Parallel execution - Multiple agents can work simultaneously on independent tasks

Expertise Separation

• Domain knowledge - Each agent has specialized instructions for its domain (Python best practices, Git workflows, Docker patterns)
• Tool restrictions - Agents only use tools relevant to their specialty
• Consistent patterns - Same agent = same coding style and conventions

Quality Assurance

• Mandatory code review - Every code change goes through 3 parallel plugin review agents
• Automated testing - test-automator runs after changes
• Review loop - Changes iterate until approved

Workflow Example

User: "Add a new feature to handle user auth"
         │
         ▼
   ┌─────────────────┐
   │  ORCHESTRATOR   │  (main Claude - manages, doesn't code)
   └────────┬────────┘
            │
   ┌────────┼────────────────┐
   │        │                │
   ▼        ▼                ▼
┌──────┐ ┌──────┐      ┌──────────┐
│python│ │ git  │      │test-auto │
│expert│ │expert│      │  mator   │
└──────┘ └──────┘      └──────────┘
   │        │                │
   └────────┴────────────────┘
            │
            ▼
  ┌──────────┬──────────┬──────────┐
  │superpow. │pr-review │feat-dev  │
  │ reviewer │ reviewer │ reviewer │
  └──────────┴──────────┴──────────┘
            │
            ▼
       Done

Benefits

• Reduced token usage - Specialist context is discarded after task completion
• Better code quality - Domain experts produce better code
• Faster execution - Parallel agents vs sequential operations
• Maintainability - Easy to add/modify agents for new domains

Orchestrator Pattern Details

The CLAUDE.md file (local-only, not tracked in git) defines an orchestrator pattern where:

1. The main Claude instance acts as a manager/orchestrator
2. It delegates tasks to specialist agents based on the domain
3. After code changes, automatic code review and test loops run

Customization

• Edit CLAUDE.md to customize orchestrator rules and agent routing
• Edit .claude/settings.json to modify hooks and allowed tools
• Edit agents in .claude/agents/ to customize agent behavior

File Structure

~/.claude/
├── agents/           # Specialist agent definitions
├── plugins/          # Plugin definitions (myk-github, myk-review, myk-qodo, myk-cursor)
├── skills/           # Skills (auto-invoked based on context)
│   └── agent-browser/
│       └── SKILL.md
├── rules/            # Orchestrator rules (auto-loaded)
├── scripts/          # Helper scripts for hooks
│   ├── git-protection.py         # Protects main branch, merged branches
│   ├── my-notifier.sh            # Custom notifications
│   ├── rule-enforcer.py          # Blocks orchestrator from using Edit/Write/Bash
│   ├── rule-injector.py          # Auto-loads rules from rules/
│   └── session-start-check.sh    # Validates required tools at session start
├── tests/            # Unit tests for Python scripts
├── settings.json     # Hooks and tool permissions
├── statusline.sh     # Status line script
├── tox.toml          # Test configuration
├── pyproject.toml    # Python project config (ruff, mypy)
├── .pre-commit-config.yaml  # Pre-commit hooks
└── LICENSE           # MIT License

Development

Running Tests

Tests are run via tox with uv:

# Run all tests
uvx --with tox-uv tox

# Run specific Python version
uvx --with tox-uv tox -e py313

# Pass pytest arguments
uvx --with tox-uv tox -- -v --tb=short

Pre-commit

This project uses pre-commit hooks for code quality:

# Install pre-commit hooks
pre-commit install

# Run on all files
pre-commit run --all-files

Contributing

When adding or modifying plugins, agents, or features, ensure all markdown documentation files in the repository are updated to reflect the changes. In particular, update AI_REVIEW.md when modifying shared project context or guidelines.

Plugins

See Quick Start (Plugins) above for installation and available commands, or plugins/README.md for detailed documentation.

Prerequisites for github/review plugins

These plugins use the myk-claude-tools CLI:

uv tool install myk-claude-tools

Or install from this repository:

uv tool install git+https://github.com/myk-org/claude-code-config

License

MIT