GitHub Copilot CLI Challenge - estazka.cz

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I built an app called eStazka. It’s a passion project designed to help my dad’s trucking business escape the paperwork. Less time shuffling papers, more time actually running the business.

Basically, it handles everything a transport company needs: transport docs ("stazka"), vehicle management, fleet tracking, and inventory. I also hooked it up to the Czech invoicing service Fakturoid, so invoices can be generated automatically from transport documents.

On top of that, it keeps an eye on vehicle status, service history, and those annoying mandatory inspections (STK).

Demo

Try it here: eStazka.cz

Demo accounts

Go ahead and log in with these credentials to see how it works:

Owner account:

• Username: [e-mail]
• Password: password

Driver account:

• Username: [e-mail]
• Password: password

My Experience with GitHub Copilot CLI

I’ve used the Copilot extension in VS Code before, but diving into the CLI for this challenge was a whole new experience.

I didn't write a single line of code manually.
The entire app was generated 100% via the GitHub Copilot CLI.

The "Zero-Code"

I tried this setup

1. The Setup: I wrote up a design doc and an app description, then fed it to the Copilot CLI to analyze and set the instructions.
2. The Plan: I lived in Plan Mode. I listed out all the requirements and features I wanted directly in the CLI.
3. The Blueprint: Once I dumped my ideas in, the CLI analyzed everything and generated a .md file for every single task in a todo folder. Think of them as AI-generated "Task Briefs."
4. The Build: I switched over to Normal Mode and told it to get to work.
5. The Loop: Copilot CLI just went through the todo folder. It implemented a feature, ran the tests, committed the changes, and moved on to the next one.

Breaking the dev process down into these micro-tasks allowed the AI to handle the heavy lifting one step at a time.

It was honestly a really fun experiment to see just how powerful these tools have become. I still love writing code myself (and I’m not stopping anytime soon), but having AI as a co-pilot—or in this case, the pilot—is something.

The AI Stack

I used a mix of models to get this done:

• Gemini 3 Pro & Claude Sonnet 4.5: These handled the creative stuff—copywriting and UI design.
• Claude Opus 4.5 & GPT 5.3 Codex: These were the workhorses for the actual feature implementation.

I was constantly toggling between Plan and Normal mode. Honestly, having that separation between "strategizing" and "executing" made the whole workflow smooth.