Skip to content

Latest commit

 

History

History
125 lines (99 loc) · 3.02 KB

README.md

File metadata and controls

125 lines (99 loc) · 3.02 KB

🏭 AI Workflows Library

Turn Complex AI Tasks into Simple, Repeatable Steps

What's This Library For?

Ever wished you could:

  • Save your best AI conversations and reuse them?
  • Share your AI expertise with your team?
  • Make sure AI gives you consistent results?
  • Build on what works instead of starting from scratch?

That's exactly what this library helps you do! Think of it as your collection of AI recipes that:

  • Work reliably every time
  • Get better with each use
  • Anyone can learn and apply
  • Scale across teams

Getting Started

1. Use an Existing Workflow

It's as simple as:

  1. Open Cursor AI's Composer (CMD/CTRL + I)
  2. Copy any run.md file content
  3. Fill in your specific details
  4. Press Enter and watch it work!

2. Create Your Own Workflow

The easiest way:

  1. Go to factory/new_workflow/
  2. Follow the steps in run.md
  3. Get a professional workflow in seconds

Library Organization

_ai_workflows/
├── code_based/            # Workflows with Python automation
│   └── faq_analysis/      # Example: FAQ improvement workflow
├── factory/               # Create new workflows
│   ├── new_workflow/      # Complete workflow template
│   └── new_instructions/  # Single task template
├── templates/             # Reusable patterns
└── text_based/            # Text-only workflows

Two Ways to Use Workflows

1. Text-Based Workflows

Perfect for when you want to:

  • Experiment quickly
  • Discuss ideas with AI
  • Refine things interactively
  • Keep things simple

2. Code-Based Workflows

Better when you need to:

  • Run things automatically
  • Have more control
  • Integrate with other tools
  • Be independent from AI assistants

Types of Workflows

Complete Workflows

For processes you'll repeat:

workflow_name/
├── instructions/    # Step-by-step guide
│   └── instructions_v1.0.0.md
└── run.md          # Ready to execute

Single Instructions

For one-time complex tasks:

project_name/
└── instructions_v1.0.0.md

Best Practices

Keep Track of Changes

  • Use version numbers (v1.0.0)
  • Document what changed
  • Keep previous versions

Ensure Quality

  • Files start with __validation_required
  • Always check the results
  • Remove the suffix after validation
  • Test with different inputs

Design Guidelines

  • One clear goal per workflow
  • Standard inputs and outputs
  • Clear success criteria
  • Regular improvements

Try an Example

Check out the FAQ analysis workflow in code_based/faq_analysis/:

  • Text version for quick experiments
  • Code version for automation
  • Complete documentation
  • Ready to use!

Learn More

Want to Contribute?

  1. Try existing workflows
  2. Test in your projects
  3. Document what you learn
  4. Share improvements
Built with ❤️ by AI Swiss