claude-code-marketplace

getty104's Claude Code Plugin Marketplace

Overview

This repository is a plugin marketplace for Claude Code that automates TDD (Test-Driven Development) based development workflows. It provides specialized agents, custom commands, and skills to streamline the entire development process, from GitHub Issue implementation to code review responses.

The marketplace format enables centralized management of multiple plugins and easy sharing with teams and communities.

Features

• TDD Automation: Automatically executes the test creation → implementation → quality check cycle
• GitHub Integration: Automates Issue implementation and review comment responses
• git worktree Utilization: Safe working environment isolated from the main branch
• Layered Architecture Compliance: Maintains consistent code structure
• Quality Assurance: Automatically runs lint, tests, and type checks
• Library Documentation Access: Integrated MCP servers for Next.js, shadcn, and other libraries

Installation

Prerequisites

• Claude Code CLI installed
• Node.js (if used in your project)
• Docker (if used in your project)
• GitHub CLI (gh)

Adding the Marketplace

Add to your Claude Code settings file (~/.config/claude/settings.json):

{
  "plugin_marketplaces": [
    "https://github.com/getty104/claude-code-marketplace"
  ]
}

Or add directly via Claude Code command:

claude marketplace add https://github.com/getty104/claude-code-marketplace

Installing Plugins

After adding the marketplace, included plugins will be automatically available:

claude plugin install getty104

MCP Configuration

The .mcp.json included in the plugin automatically configures the following MCP servers:

• chrome-devtools: Browser automation and DevTools integration
• context7: Library documentation retrieval (HTTP-based, no API key required)
• next-devtools: Next.js development tools and documentation
• shadcn: shadcn/ui component library integration

Marketplace Composition

This marketplace includes the following plugins:

getty104 Plugin

An integrated plugin for automating TDD development workflows. Composed of five component types:

1. Agents (agents/) - Specialized sub-agents
2. Commands (commands/) - Custom slash commands
3. Skills (skills/) - Reusable skill prompts
4. Hooks (hooks/hooks.json) - Event handlers
5. MCP Servers (.mcp.json) - External tool integration

Key Features

Agents

github-issue-implementer

Specialized agent for implementing GitHub Issues and creating PRs

Features:

• Analysis of Issue content and implementation planning
• Implementation following TDD cycle
• Code quality checks (lint, tests)
• Automatic PR creation

Usage Example:

Implement this Issue #123 and create a PR

review-comment-implementer

Agent specialized in implementing review comments

Features:

• Retrieval of unresolved review comments
• TDD approach for fixes
• Resolving review comments
• Automation of re-review requests

Usage Example:

Check PR review comments and fix the pointed out issues

task-requirement-analyzer

Agent specialized in analyzing task requirements and creating implementation plans

Features:

• Deep analysis of task requirements (explicit and implicit)
• TDD-based phased implementation planning
• Risk and concern identification

Usage Example:

Analyze the requirements for adding password reset functionality

general-purpose-assistant

General-purpose agent for diverse tasks requiring broad problem-solving capabilities

Features:

• Comprehensive problem analysis and solution
• Adherence to project conventions (TDD, no comments, layered architecture)
• Integration with Context7 MCP
• Flexible task execution across multiple domains

Usage Example:

Explain the overall project structure
Provide advice on improving development efficiency

Slash Commands

/exec-issue <issue number>

Reads GitHub Issue and automates from implementation to PR creation

Execution Steps:

1. Create git worktree
2. Analyze Issue content
3. TDD implementation
4. PR creation

/create-worktree <branch name>

Creates and sets up a git worktree for task execution

Execution Steps:

1. Create git worktree with specified branch name
2. Copy environment files (.env)
3. Install dependencies

/create-plan <task description>

Creates an implementation plan and GitHub Issue using task-requirement-analyzer

Execution Steps:

1. Move to default branch and pull latest changes
2. Analyze task requirements using task-requirement-analyzer agent
3. Create GitHub Issue with implementation plan

/fix-review-point <branch name>

Address unresolved review comments on specified branch

Execution Steps:

1. Prepare git worktree
2. Retrieve review comments
3. TDD fixes
4. Resolve comments
5. Re-review request

/fix-review-point-loop <branch name>

Repeatedly address review comments until none remain (checks every 5 minutes)

/general-task <task description>

Execute general tasks using the general-purpose-assistant agent

Execution Steps:

1. Analyze task requirements
2. Select appropriate approach and tools
3. Execute task with Context7 MCP as needed
4. Validate results against project standards

Skills

Skills are reusable prompts that can be invoked to perform specific tasks. Each skill can be invoked using the /skill-name syntax (e.g., /check-library, /web-search).

/check-library

Retrieves library documentation using appropriate MCP servers

Features:

• Next.js: Uses next-devtools MCP
• shadcn/ui: Uses shadcn MCP
• Other libraries: Uses context7 MCP

Usage Example:

/check-library React Query

/create-git-worktree

Automates git worktree creation and environment setup

Features:

• Creates worktree in .git-worktrees/ directory
• Automatically converts / to - in branch names
• Copies .env files
• Installs dependencies (npm install)
• Reuses existing worktrees

Usage Example:

/create-git-worktree feature/new-feature

/web-search

Advanced web search using gemini command. Preferred over default WebSearch tool for comprehensive searches.

Features:

• Natural language queries
• Complex question handling
• Source citation
• Multi-result integration

Usage Example:

/web-search How to implement authentication with Next.js App Router

/high-quality-commit

Guides appropriate git commit strategies for code changes

Features:

• Squash strategy (default): Amend to existing commits
• New commit: For independent changes
• Interactive rebase: For reorganizing commit history
• Commit message guidelines (type, subject, body, footer)

Usage Example:

/high-quality-commit

/read-unresolved-pr-comments

Retrieves unresolved PR comments via GitHub GraphQL API

Features:

• Fetches unresolved Review threads (code-specific comments, resolvable)
• Fetches Issue comments with code blocks (conversation tab, not resolvable)
• Returns PR metadata (number, title, URL, state, author, reviewers)
• JSON output format

Usage Example:

/read-unresolved-pr-comments

/resolve-pr-comments

Batch resolves PR Review threads via GitHub GraphQL API

Features:

• Automatically resolves all unresolved Review threads
• Uses resolveReviewThread mutation
• Displays resolve results for each thread
• Note: Issue comments (conversation tab) are not resolvable

Usage Example:

/resolve-pr-comments

Development Guidelines

TDD (Test-Driven Development)

1. Create tests (in same directory as test target file)
2. Run tests (confirm failure)
3. Implementation
4. Run tests (confirm success)
5. Refactoring

Code Quality Standards

• All tests must pass
• Zero errors from npm run lint
• TypeScript type safety ensured
• No comments (code should be self-explanatory)
• Minimal file changes

Layered Architecture

• Model Layer: Business logic
• Infrastructure Layer: Database, external APIs
• Application Layer: Use cases
• Presentation Layer: UI/API responses

Directory Structure

claude-code-marketplace/
├── .claude-plugin/
│   └── marketplace.json                    # Marketplace manifest
│
├── getty104/                               # getty104 plugin
│   ├── .claude-plugin/
│   │   └── plugin.json                     # Plugin manifest (auto-generated)
│   ├── .mcp.json                           # MCP server configuration
│   ├── agents/                             # Agent definitions
│   │   ├── github-issue-implementer.md     # Issue implementation agent
│   │   ├── review-comment-implementer.md   # Review response agent
│   │   ├── task-requirement-analyzer.md    # Task analysis agent
│   │   └── general-purpose-assistant.md    # General-purpose agent
│   ├── commands/                           # Command definitions
│   │   ├── exec-issue.md                   # Issue implementation command
│   │   ├── create-worktree.md              # Worktree creation command
│   │   ├── create-plan.md                  # Implementation plan command
│   │   ├── fix-review-point.md             # Review response command
│   │   ├── fix-review-point-loop.md        # Full review response command
│   │   └── general-task.md                 # General task execution command
│   ├── skills/                             # Skill definitions
│   │   ├── check-library/                  # Library documentation skill
│   │   │   ├── SKILL.md
│   │   │   └── examples.md
│   │   ├── create-git-worktree/            # Worktree creation skill
│   │   │   ├── SKILL.md
│   │   │   └── scripts/
│   │   │       └── create-worktree.sh
│   │   ├── web-search/                     # Web search skill
│   │   │   ├── SKILL.md
│   │   │   ├── examples.md
│   │   │   └── scripts/
│   │   │       └── web-search.sh
│   │   ├── high-quality-commit/            # Commit strategy skill
│   │   │   ├── SKILL.md
│   │   │   ├── examples.md
│   │   │   └── reference.md
│   │   ├── read-unresolved-pr-comments/    # PR comment retrieval skill
│   │   │   ├── SKILL.md
│   │   │   └── scripts/
│   │   │       └── read-unresolved-pr-comments.sh
│   │   └── resolve-pr-comments/            # PR comment resolution skill
│   │       ├── SKILL.md
│   │       └── scripts/
│   │           └── resolve-pr-comments.sh
│   ├── scripts/                            # Utility scripts
│   │   └── remove-merged-worktrees.sh      # Cleanup merged worktrees
│   └── hooks/                              # Hook definitions
│       └── hooks.json                      # Event handlers
│
├── CLAUDE.md                               # Claude Code guide
└── README.md                               # This file

Marketplace Manifest

.claude-plugin/marketplace.json defines marketplace metadata:

{
  "name": "marketplace",
  "metadata": {
    "description": "getty104's marketplace",
    "version": "0.0.1"
  },
  "owner": {
    "name": "getty104"
  },
  "plugins": [
    {
      "name": "getty104",
      "source": "./getty104",
      "description": "This is a getty104's base tool set"
    }
  ]
}

Development and Customization

Adding Agents

Create .md files in getty104/agents/. Define name, description, model, and color in frontmatter:

---
name: custom-agent
description: Description of when this agent should be used (Claude Code uses for auto-judgment)
model: sonnet
color: cyan
---

Write agent prompt content here.
This prompt will be passed to the sub-agent.

Best Practices:

• Make description clear and specific (Claude Code automatically invokes at appropriate times)
• Create agents with specialized roles
• Specify communication requirements (e.g., Japanese language)

Adding Commands

Create .md files in getty104/commands/ and describe the processing content:

---
allowed-tools: Bash(git *), Context7(*)
description: Description of the command
---

Describe task description and execution steps in Markdown format.

## Step 1
Processing content...

## Step 2
Processing content...

Commands are executed as /command-name arguments, and can reference arguments via the $ARGUMENTS variable.

Adding Skills

Create a directory in getty104/skills/ with a SKILL.md file:

---
name: skill-name
description: Description of what this skill does and when to use it
---

# Skill Title

## Instructions

Detailed instructions for executing the skill...

Skills can include the following structure:

skills/
└── my-skill/
    ├── SKILL.md              # Main skill definition (required)
    ├── scripts/              # Shell scripts for automation (optional)
    │   └── my-script.sh
    ├── examples.md           # Usage examples (optional)
    └── reference.md          # Additional reference material (optional)

Best Practices:

• Make description clear and specific for auto-invocation by Claude Code
• Include ## Instructions section with step-by-step guidance
• Place shell scripts in scripts/ subdirectory for organization
• Reference scripts using relative paths: bash scripts/my-script.sh

Configuring Hooks

Configure event handlers in getty104/hooks/hooks.json:

{
  "hooks": {
    "Stop": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "your-command-here"
          }
        ]
      }
    ]
  }
}

Available events: PreToolUse, PostToolUse, Stop, etc.

Adding MCP Servers

Add new MCP servers to .mcp.json:

{
  "mcpServers": {
    "server-name": {
      "command": "command",
      "args": ["arg1", "arg2"]
    }
  }
}

For HTTP-based MCP servers:

{
  "mcpServers": {
    "server-name": {
      "type": "http",
      "url": "https://example.com/mcp"
    }
  }
}

Plugin Validation

Always validate plugins before distribution:

claude plugin validate getty104/

Run local tests:

claude plugin install ./getty104

Sharing the Marketplace

To share this marketplace with teams or communities:

Public Sharing

Publish repository on GitHub and share the URL:

https://github.com/getty104/claude-code-marketplace

Private Sharing

Grant team members access to the repository, then use the same URL.

Fork and Customize

Fork this marketplace to add your own plugins:

1. Fork the repository
2. Update owner in .claude-plugin/marketplace.json
3. Add new plugins or modify existing ones
4. Validate and commit

Troubleshooting

Plugin Not Recognized

Check with:

claude plugin list

Verify that the marketplace has been added correctly.

MCP Server Won't Start

Ensure the required tools are installed:

• npx for Node.js-based servers
• uvx for Python-based servers (requires uv)

Command Won't Execute

Verify that command file names are correct. File names become command names as-is.

Resources

• Claude Code Official Documentation
• Plugin Marketplace Guide
• Plugin Reference

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss the proposed changes.

License

MIT

Author

getty104

Support

For questions or issues, please report them in GitHub Issues.