Supercharging Your Projects with Cursor AI: The Instructions Generator Approach

In the rapidly evolving landscape of AI-assisted development, I’ve discovered a game-changing approach that has transformed how I start new projects. By creating a meta-project called the “Project Instructions Generator,” I’ve been able to systematically harness the power of Cursor AI to bring ideas to life with remarkable efficiency. Let me share my journey and how you might benefit from a similar approach.

The Problem: Starting from Scratch

Every new project begins with a blank slate, an empty folder, a flashing cursor, and a head full of ideas. But translating those ideas into a structured, well-organized project can be daunting. Even with powerful AI assistants like Cursor’s Agent, there’s often a disconnect between our vision and the AI’s ability to understand and execute on that vision.

I found myself spending significant time in the early stages of projects just trying to get the AI to understand my goals, establish the right structure, and set up consistent workflows. The instructions I gave were often ad-hoc, inconsistent, and lacked the necessary detail to guide the AI effectively.

The Solution: The Instructions Generator Approach

The breakthrough came when I realized that the most effective way to start a new project with Cursor AI was to begin with a comprehensive set of instructions, a “living document” that would guide both myself and the AI throughout the project lifecycle.

Enter the Project Instructions Generator, a meta-project designed to create detailed, customized instructions for any type of project I might want to undertake.

How It Works

The system is elegantly simple:

I open the Agent chat in my Instructions Generator project
I describe my new project idea in detail
The Agent generates a comprehensive set of instructions as a markdown file
I copy those instructions to my new project folder
I attach the instructions file to the Agent chat in the new project
I tell Cursor AI to follow the instructions
The magic happens, the Agent systematically builds out my project following the roadmap we’ve established

The Power of Structured Instructions

The generated instructions aren’t just simple to-do lists. They’re comprehensive documents that cover the entire project lifecycle:

Project Overview - Clear definition of goals, audience, and success criteria
Initial Planning - Resource needs, dependencies, and research requirements
Project Plan - Detailed phases with checklist-style tasks
Setup Guidelines - Environment configuration and folder structure
Workflow Management - Documentation practices, version control, and progress tracking
Quality Control - Review processes and testing methodologies
Completion & Delivery - How to finalize and share the project
Documentation - Standards for documenting the project and its processes

Projects I’ve built… so far…

This approach has revolutionized how I work with Cursor AI across various project types. Here are some of the projects I’ve successfully developed using this method:

Solo Leveling Quest Generator

My first test of this method was creating a daily quest generator app inspired by the leveling system from the anime “Solo Leveling.” The application was supported by local LLM calls via Ollama to Llama 3.1 and DeepseekR1.

The initial results were quite promising, though I did encounter some limitations in the overall project development. I’m planning to restart this project soon using my refined process, which should help overcome the previous challenges.

Homelab Idle Clicker Game

I built a fun, small idle clicker client/server game as a sort of “hello world” installation for a new homelab server I’ve been experimenting with. It served as both a learning experience and a practical application of the instructions generator approach.

I’ll be sharing more details in an upcoming blog article as I continue to explore more homelab experiments!

”Update Protocol” - A Sci-Fi Novel

Perhaps my most ambitious project was writing and revising an entire sci-fi book exploring the themes of machine sentience and consciousness. Titled “Update Protocol,” I managed to complete both the first and second drafts in a single weekend thanks to the structured approach provided by the instructions generator.

You can read more about the book in its dedicated blog post here or explore the full book online here, where I’ve made it available for feedback.

Shower Thoughts to Project Plans

Beyond creating new projects from scratch, I’ve been systematically converting old “shower thoughts” ideas from my personal notebooks and journals into this structured format. This has allowed me to build up a collection of project plans that are ready to implement whenever I have free time. (and, of course, monthly fast requests available)

While this isn’t exactly creating new projects, it provides me with a solid foundation to work on side projects during my limited free time, projects that I ordinarily wouldn’t even have time to plan, let alone make progress on.

If you’re interested in trying this approach yourself, I’ve published a starter repository at https://github.com/adam-lohnes/instructions-generator to help you get started. It’s obviously an ongoing work in progress, but it’s in a pretty solid working state, I think.

The Secret Ingredient: Progress Journals

One of the most valuable elements I’ve built into the instruction sets is the concept of progress journals. For each project, the Agent:

Creates a progress-journal/ directory
Checks the current date via the terminal
Creates dated markdown files for each feature or milestone
Updates these journals with detailed information about what was accomplished

This practice has provided incredible visibility into project development. I can see exactly what was done, when, and why, making it easier to pick up projects after breaks or share progress with collaborators.

Why This Approach Works So Well

The Instructions Generator approach succeeds for several reasons:

Consistency - The AI follows a consistent pattern across all projects
Comprehensiveness - Nothing falls through the cracks because everything is documented
Adaptability - Instructions evolve as the project does
Clarity - Both you and the AI understand exactly what’s expected
Efficiency - Less time spent explaining what you want, more time spent building it

Future Potential with MCP Integration

As Multi-model Collaboration Protocol (MCP) becomes more widely adopted, this approach will become even more powerful. The instructions can include guidance for integration with external tools, APIs, and services, allowing the Agent to:

Pull in design assets from Figma
Integrate with project management tools
Analyze data from various sources
Deploy to production environments
Perform user testing and collect feedback

Getting Started with Your Own Instructions Generator

If you’d like to implement a similar approach, you can check out my starter repository at https://github.com/adam-lohnes/instructions-generator, or:

Create a .cursor/rules/ directory in your project
Add specialized rules for different project types
Organize template examples in a structured directory
Create a README explaining how to use the system

Remember that Cursor rules should be placed directly in the .cursor/rules/ directory without subdirectories, and you may want to use prefixes (like tech_, creative_, etc.) to keep them organized.

Conclusion

The Project Instructions Generator approach has fundamentally changed how I interact with AI assistants in my creative and development work. By starting with clear, comprehensive instructions, I’ve been able to harness the full potential of Cursor AI, resulting in more efficient workflows, higher quality output, and a more enjoyable development experience.

The key insight is that AI assistants, no matter how advanced, benefit tremendously from structured guidance. By investing time upfront to create detailed instructions, you’ll save countless hours throughout the project lifecycle while achieving more consistent and impressive results.

I’m excited to see how others might adapt and improve upon this approach as AI-assisted development continues to evolve. If you try this method with the instructions generator repository, I’d love to hear about your experiences and any enhancements you discover along the way!

What project will you build next with your AI partner? With the right instructions, the possibilities are limitless.