minimal-claude

Minimal Claude Code Plugin

A minimal base setup for creating Claude Code plugins with an intelligent /setup command that automatically configures project linting and typechecking.

Structure

minimal-claude/
├── .claude-plugin/
│   └── plugin.json          # Plugin manifest (required)
├── commands/                # Slash commands
│   ├── setup-code-quality.md # Generate /fix command
│   ├── setup-claude-md.md    # Generate CLAUDE.md guidelines
│   ├── setup-commits.md      # Generate /commit command
│   └── example.md
├── agents/                  # Subagents
│   └── example-agent.md
├── skills/                  # Agent Skills
│   └── SKILL.md
└── hooks/                   # Event hooks
    └── hooks.json

Featured Command: /setup-code-quality

The /setup-code-quality command intelligently detects your project type and configures automated code quality checks:

1. Detects Project Type: Automatically identifies if you're using JavaScript/TypeScript, Python, Go, Rust, PHP, or Java
2. Checks Existing Tools: Verifies which linting and typechecking tools are already installed
3. Installs Missing Tools: Only installs what's needed (eslint, prettier, mypy, clippy, etc.)
4. Generates /fix Command: Creates a custom /fix command tailored to your project that:
   • Runs all your linting and typechecking tools
   • Parses errors and groups them by domain (types, lint, formatting)
   • Spawns parallel agents to fix all issues simultaneously

Usage

/setup-code-quality

After running setup, use the generated command:

/fix

This will automatically fix all linting and type errors in your project using parallel agents!

Featured Command: /setup-claude-md

The /setup-claude-md command creates a minimal, non-bloated CLAUDE.md file with zero-tolerance code quality guidelines:

1. Detects Project Type: Identifies your project and its tooling
2. Extracts Exact Commands: Finds the actual lint/typecheck commands from your config
3. Generates CLAUDE.md: Creates a minimal guidelines file (under 100 lines) that:
   • Enforces running checks after EVERY file edit
   • Implements zero-tolerance for errors/warnings
   • Uses your project's specific commands
   • Gets automatically injected into Claude's prompt

Usage

/setup-claude-md

This creates a CLAUDE.md file that tells Claude to automatically run your linting and typechecking commands after every edit, fixing all issues immediately.

Example generated CLAUDE.md (for a TypeScript project):

# Code Quality Guidelines

**Zero-tolerance policy**: All code must pass linting and type checking.

## After Every Edit

After writing, editing, or updating ANY file, you MUST:

1. Run these commands:
   ```bash
   npm run lint
   npm run typecheck

2. Fix ALL errors and warnings immediately
3. Re-run checks until zero errors/warnings remain

**Minimal, actionable, effective.**

## Featured Command: `/setup-commits`

The `/setup-commits` command creates a `/commit` command that enforces quality checks before committing:

1. **Detects Project Type**: Identifies your project's linting/typechecking tools
2. **Runs Quality Checks**: Executes all checks before allowing commits
3. **Generates Smart Commit Messages**: AI-powered, human-readable commit messages
4. **Auto-push**: Automatically pushes to remote after committing

### Usage

```bash
/setup-commits

This creates a /commit command that:

1. Runs npm run lint and npm run typecheck (or your project's equivalent)
2. Only proceeds if all checks pass (zero tolerance)
3. Analyzes your changes
4. Generates a clear commit message like "Add user authentication with JWT"
5. Commits and pushes automatically

Example workflow:

# Make changes to your code
# Then run:
/commit

# Claude will:
# ✓ Run all quality checks
# ✓ Generate commit message
# ✓ Commit changes
# ✓ Push to remote

Plugin Components

Commands (/commands)

Custom slash commands that users can invoke. Each command is a Markdown file with frontmatter:

---
description: Command description
---

Your command prompt here.

Agents (/agents)

Specialized subagents that Claude can invoke for specific tasks. Define agents in Markdown files with frontmatter.

Skills (/skills)

Agent Skills that Claude can invoke autonomously based on context. Skills are defined in SKILL.md files.

Hooks (/hooks)

Event handlers that respond to Claude Code actions. Configure in hooks.json:

• PostToolUse - After any tool is used
• UserPromptSubmit - When user submits a prompt
• SessionStart - When a session starts

Installation

Option 1: Install from GitHub (Recommended)

Once you've pushed this to GitHub, users can install it directly:

# Add the marketplace
/plugin marketplace add your-username/minimal-claude

# Install the plugin
/plugin install minimal-claude@your-username