PromptyDumpty

<img src="website/public/logo.png" width=200px />

A lightweight, universal package manager for AI coding assistants (prompts, instructions, rules, workflows, etc.).

🌐 Visit dumpty.dev for full documentation and guides.

What is it?

PromptyDumpty lets you install and manage prompt packages across different AI coding assistants like GitHub Copilot, Claude, Cursor, Gemini, Windsurf, OpenCode, and more.

Why PromptyDumpty?

The Problem

Without a package manager, sharing AI assistant configurations is painful:

• 📋 Manual file management - Copying .md files between projects and agent folders
• 🔄 No updates - When prompts improve, you manually re-copy everywhere
• 🗑️ Messy removal - Deleting prompts leaves orphaned files scattered across directories
• 🤷 No tracking - Can't remember what's installed or where it came from
• 🔁 Multi-agent duplication - Maintaining separate versions for Copilot, Claude, Cursor, etc.
• 📧 Ad-hoc sharing - Sending via email/Slack with instructions like "just copy to .claude/commands/"

The Solution

• Share prompts easily: Package and distribute your team's prompts via Git
• Works everywhere: One package works with multiple AI coding assistants
• Simple: Just YAML files and Git repos, no complex setup
• Safe: Clean installation and removal, clear tracking with lockfiles

Quick Start

# Initialize in your project
dumpty init

# Install a package
dumpty install https://github.com/org/my-prompts

# List installed packages
dumpty list

# Update packages
dumpty update --all

# Remove a package
dumpty uninstall my-prompts

How it works

1. Auto-detects your AI agent (checks for .github/prompts/, .claude/commands/, etc.)
2. Installs package files to the right directories
3. Tracks everything in a lockfile for easy management
4. Organizes files by package name for clean removal

Package Structure

Organize your files however you want! The manifest defines everything:

my-package/
├── dumpty.package.yaml  # Package manifest
├── README.md
└── src/                 # Any structure you prefer
    ├── planning.md
    ├── review.md
    └── standards.md

Creating Packages

Define what your package provides in dumpty.package.yaml - organized by agent and type:

name: my-workflows
version: 1.0.0
description: Custom development workflows
manifest_version: 1.0
author: Your Name
license: MIT
homepage: https://github.com/org/my-workflows

agents:
  copilot:
    prompts:
      - name: code-review
        description: Code review workflow
        file: src/review.md
        installed_path: code-review.prompt.md
    agents:
      - name: standards
        description: Coding standards agent
        file: src/standards.md
        installed_path: standards.agent.md
    instructions:
      - name: repo-guidelines
        description: Repository-specific coding guidelines
        file: src/guidelines.md
        installed_path: guidelines.instructions.md

  claude:
    commands:
      - name: code-review
        description: Code review command
        file: src/review.md
        installed_path: review.md

Key Features:

• Organize files however makes sense to you
• Organize artifacts by type (prompts, agents, rules, commands, etc.)
• Use "files" type for generic artifacts
• Explicitly map each file to its install location per agent
• Reuse the same source file for multiple agents
• Installation paths: {agent_dir}/{type}/{package}/{file}

External Repository References

Reference files from external repositories without forking.

Add external_repository to your manifest to pull files from another repository:

name: curated-prompts
version: 1.0.0
manifest_version: 1.0
author: Your Name
license: MIT
external_repository: https://github.com/community/prompts@a1b2c3d4e5f6789012345678901234567890abcd

agents:
  copilot:
    prompts:
      - name: refactoring
        file: prompts/refactoring.md
        installed_path: refactoring.prompt.md

Format: <git-url>@<40-char-commit-hash>

When external repo is specified:

• All file paths resolve from the external repository
• Both repos are tracked in lockfile
• Use full commit hashes only (no tags/branches)

Use cases:

• Curate content from large community repositories
• Version-lock third-party prompts
• Create team-specific views of shared content

Documentation

📚 Full documentation available at dumpty.dev

• Getting Started Guide
• Creating Packages
• Full Documentation

Installation

From Source

# Clone the repository
git clone https://github.com/dasiths/PromptyDumpty.git
cd PromptyDumpty

# Install in development mode (recommended for contributors)
make install-dev

# Or install in production mode
make install

Using pip (when published)

pip install prompty-dumpty

Verify Installation

dumpty --version

Development

Prerequisites

• Python 3.9 or higher
• Git
• Make (optional, for using Makefile commands)

Setup Development Environment

# Clone and navigate to repository
git clone https://github.com/dasiths/PromptyDumpty.git
cd PromptyDumpty

# Install in development mode with all dependencies
make install-dev

Available Make Commands

Python/CLI Commands:

make help          # Show all available commands
make test          # Run tests
make test-cov      # Run tests with coverage report
make lint          # Run linters (ruff and black)
make format        # Format code with black
make build         # Build distribution packages
make clean         # Remove build artifacts
make run ARGS='...'  # Run dumpty CLI

Website Commands:

make website-install  # Install website dependencies
make website-dev      # Start dev server with hot reload
make website-build    # Build website for production
make website-preview  # Preview production build
make website-clean    # Remove website build artifacts

Running Tests

# Run all tests
make test

# Run tests with coverage
make test-cov

# Run specific test file
pytest tests/test_models.py -v

Code Quality

# Check code formatting and linting
make lint

# Auto-format code
make format

Usage Examples

Initialize a Project

# Auto-detect agents in current directory
dumpty init

# Initialize with specific agent
dumpty init --agent copilot
dumpty init --agent claude

Install Packages

# Install from GitHub repository
dumpty install https://github.com/org/my-prompts

# Install specific version tag
dumpty install https://github.com/org/my-prompts --version 1.0.0

# Install for specific agent
dumpty install https://github.com/org/my-prompts --agent copilot

List Installed Packages

# Show installed packages (table view)
dumpty list

# Show detailed information
dumpty list --verbose

Using the Makefile

# Run CLI commands using make
make run ARGS='--version'
make run ARGS='init --agent copilot'
make run ARGS='list'
make run ARGS='install https://github.com/org/my-prompts'

Supported AI Coding Assistants

• GitHub Copilot (.github/)
• Claude (.claude/)
• Cursor (.cursor/)
• Gemini (.gemini/)
• Windsurf (.windsurf/)
• Cline (.cline/)
• Aider (.aider/)
• Continue (.continue/)
• OpenCode (.opencode/)

Project Structure

PromptyDumpty/
├── dumpty/              # Main package
│   ├── cli.py          # CLI entry point
│   ├── models.py       # Data models
│   ├── agent_detector.py  # Agent detection
│   ├── downloader.py   # Package downloading
│   ├── installer.py    # File installation
│   ├── lockfile.py     # Lockfile management
│   └── utils.py        # Utilities
├── tests/              # Test suite
├── website/            # Documentation website (dumpty.dev)
│   ├── src/           # React source files
│   ├── public/        # Static assets
│   └── README.md      # Website development guide
├── docs/              # Documentation and planning
├── examples/          # Example packages and demos
├── pyproject.toml     # Project configuration
├── Makefile          # Build and development tasks
└── README.md         # This file

Website Development

The project website is built with Vite + React and deployed at dumpty.dev.

Running the Website Locally

# Install dependencies
make website-install

# Start dev server (with hot reload)
make website-dev

Visit http://localhost:5173 in your browser.

Building for Production

# Build the website
make website-build

# Preview production build
make website-preview

See website/README.md for more details and website/DEPLOYMENT.md for deployment instructions.

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run tests: make test
5. Format code: make format
6. Check linting: make lint
7. Submit a pull request

License

MIT