DEV Community: Vadim

Copilot CLI helped me with VS Code accessibility extension

Vadim — Sat, 14 Feb 2026 19:35:22 +0000

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I created Axe Accessibility Checker — a VS Code extension powered by axe-core that runs accessibility audits directly in VS Code. Built entirely with GitHub Copilot CLI assistance, this extension shortens the gap between accessibility testing and AI-powered helpers for fixing.

Key Features:

🔍 Instant accessibility audits using industry-standard axe-core
📊 Beautiful HTML reports with violation categorization (critical, serious, moderate, minor)
⚡ Integrated with VS Code's command palette for seamless workflow
🤖 Reports optimized for GitHub Copilot consumption — copy violations directly into chat for AI-powered fixes

Why I Built It

Accessibility testing shouldn't feel like something difficult . The traditional workflow — run Lighthouse, export report, copy-paste violations into Copilot, explain context — wastes precious development time.

This extension eliminates that friction entirely. Run an audit, click to open the report, and feed violations directly to GitHub Copilot for instant fix suggestions. What used to take 5-10 minutes now takes seconds.

Personal Journey: I'd never created a VS Code extension before this challenge. GitHub Copilot CLI walked me through the entire process — from manifest configuration to Puppeteer integration to TypeScript compilation. This extension is proof that Copilot CLI can guide you through completely unfamiliar territory.

I'm seriously considering publishing this on the VS Code Marketplace because it solves a real problem I face daily.

Demo

🎥 Watch the Demo

Repository

💻 Project Repository

🔌 VS Code Extension link

🚧 WIP — coming soon™

My Experience with GitHub Copilot CLI

What I Loved:

/delegate [prompt] — The killer feature! Send your entire session to GitHub and Copilot creates a PR automatically. This is game-changing for collaborative workflows.
/on-air — Streamer mode that hides model names and quotas. Perfect for demos and recordings without exposing implementation details.
/tasks — Background task management for subagents and shell sessions. Feels like having a real task manager for AI work.
Context awareness — Copilot CLI understood my project structure and suggested changes across multiple files simultaneously.

What Could Be Better:

Fleet mode unpredictability — I asked it to find code issues and propose improvements. After a while, it started making edits without clear explanations of the benefit. Felt like it was "moving fast and breaking things" without asking permission.
Silent operations — Sometimes goes quiet for 60-200 seconds with no indication of what it's processing. A simple "analyzing codebase..." would help.
Terminal session handling — Occasionally lost track of which terminal was active when switching between compile and test sessions.

Overall: Despite minor quirks, GitHub Copilot CLI elevated my productivity significantly. This project would've taken me 3-7 days alone — Copilot CLI got me there in hours.

Team

Solo Developer: Vadim Frolov

Designer • Developer • Producer

(With a little help from my AI friend)

How I Built a 69KB Portfolio with Google's Coolest Tools

Vadim — Sat, 24 Jan 2026 19:20:56 +0000

About Me

I’m a Software Engineer with 7+ years of experience, previously working as a Civil Engineer — which heavily influenced how I think about structure, performance, and long-living systems.

I live in Germany and can speak a little bit Chinese.🍊

Portfolio

The portfolio focuses on minimal JavaScript, clean UI and fast load times.

https://vadim-frolov-gemini-hack-bgu43ly7xa-uc.a.run.app

How I Built It

I built my portfolio with Antigravity, Gemini CLI, HTML, CSS and JavaScript. I was looking for retro style simple portfolio with keyboard navigation

Steps

I've started with planning mode and simple prompt, I'm big fan of spec-driven development
I've checked implementation plan and add couple of manual changes and proceed with it
With built-in browser I've easily checked visuals
During implementing some small tweaks my tokens were all spent so I install Gemini CLI and finished small CSS changes

Challenges & Trade-offs

Keeping accessibility while maintaining a retro look
Avoiding over-engineering when using AI tools
Balancing minimal CSS with responsive behavior

What I'm Most Proud Of

I'm proud to create light-weight website, only 69kb, that runs pretty well on desktop and mobile. It was fun to use Google Tools such as Gemini CLI and Antigravity.

I also uploaded it Google Cloud, it was my first try with this service, it pretty straightforward experience, thought I use some help of Gemini to advance in couple of steps.

I think it's important to think about performance and UX, I want to create something that will not need maintenance in the future and will be fun to use.

I'm glad of my result and had a good time building it.

Feedback is welcome — especially from people who’ve tried similar retro designs.

Why Kiro IDE Made Me Fall in Love with AI-Assisted Development

Vadim — Sat, 13 Sep 2025 18:53:12 +0000

I'll be honest - I was skeptical about AI coding tools. But after building my first project with Kiro IDE, I'm completely converted. Here's why you should give it a shot.

The Magic of Automated Spec Flow

Picture this: You have a project idea, and instead of staring at a blank screen wondering where to start, Kiro takes you through this incredible flow:

Requirements → Design Docs → Task List → Code

All automated. All intelligent. All surprisingly accurate.

I recently built a local-first smart planner using React 19, FastAPI, and PostgreSQL. What would normally take me hours of planning and architecting happened in minutes.

What Makes Kiro Different

🧠 It Actually Understands Context

Unlike other AI tools that give you random code snippets, Kiro maintains context across your entire project. It knows your architecture, your patterns, and your preferences.

📋 Comprehensive Planning

Before writing a single line of code, Kiro generates:

Detailed requirements analysis
System architecture documents
Component breakdowns
Structured task lists

🏗️ Real Development Flow

It doesn't just generate code - it guides you through proper development practices while actually building features that work together.

My Experience: From Idea to Working App

Starting with just "I want a local-first todo app with AI features," Kiro helped me create:

✅ Complete React 19 frontend with TypeScript
✅ FastAPI backend with async patterns
✅ PostgreSQL database models
✅ Ollama integration for AI task generation
✅ Real-time chat interface

The development felt effortless. Not because it was doing everything for me, but because it eliminated all the friction and decision paralysis that usually slows me down.

Why You Should Try It

If you're a beginner: Kiro teaches you best practices while you build real projects.

If you're experienced: It handles the boring boilerplate and architecture decisions so you can focus on the creative parts.

If you're curious: It's genuinely fun to see how AI can enhance (not replace) your development workflow.

The Bottom Line

I'm already planning my second Kiro project. When was the last time a dev tool made you excited to code more?

Try it out and let me know what you build! Drop your projects in the comments - I'd love to see what the community creates.

Have you tried AI-assisted development? What's been your experience? Let's discuss in the comments!

Demo reel

Why Kiro IDE Made Me Fall in Love with AI-Assisted Development

Vadim — Sat, 13 Sep 2025 18:48:06 +0000

I'll be honest - I was skeptical about AI coding tools. But after building my first project with Kiro IDE, I'm completely converted. Here's why you should give it a shot.

The Magic of Automated Spec Flow

Picture this: You have a project idea, and instead of staring at a blank screen wondering where to start, Kiro takes you through this incredible flow:

Requirements → Design Docs → Task List → Code

All automated. All intelligent. All surprisingly accurate.

I recently built a local-first smart planner using React 19, FastAPI, and PostgreSQL. What would normally take me hours of planning and architecting happened in minutes.

What Makes Kiro Different

📋 Comprehensive Planning

Before writing a single line of code, Kiro generates:

Detailed requirements analysis
System architecture documents
Component breakdowns
Structured task lists

🏗️ Real Development Flow

It doesn't just generate code - it guides you through proper development practices while actually building features that work together.

My Experience: From Idea to Working App

Starting with just "I want a local-first todo app with AI features," Kiro helped me create:

✅ Complete React 19 frontend with TypeScript
✅ FastAPI backend with async patterns
✅ PostgreSQL database models
✅ Ollama integration for AI task generation
✅ Real-time chat interface

The development felt effortless. Not because it was doing everything for me, but because it eliminated all the friction and decision paralysis that usually slows me down.

Why You Should Try It

If you're a beginner: Kiro teaches you best practices while you build real projects.

If you're experienced: It handles the boring boilerplate and architecture decisions so you can focus on the creative parts.

If you're curious: It's genuinely fun to see how AI can enhance (not replace) your development workflow.

The Bottom Line

I'm already planning my second Kiro project. When was the last time a dev tool made you excited to code more?

Try it out and let me know what you build! Drop your projects in the comments - I'd love to see what the community creates.

Have you tried AI-assisted development? What's been your experience? Let's discuss in the comments!

Demo reel

Cool GitHub Copilot feature in VS Code you may not have heard of

Vadim — Sun, 27 Jul 2025 15:55:44 +0000

Have you ever wished someone check your code and provide feedback before you even submit a commit?

I discovered GitHub Copilot Code Review for uncommitted changes — and I've immediately started using it.

How It Works

Each time you finish the piece of work you can ask for check and get some useful advices even before it will be submitted to repository.

If you click on the button with chat icon in "Source Control" section you can get the same experience from Github Copilot Review - but locally right inside IDE.

What It Can Do

This feature can help you:

Quickly fix variable names for better clarity
Refactor overcomplicated code into cleaner solutions
Propose additional checks for edge cases you might have missed

Why I♥️it

I find this feature extremely useful because it can prevent extra comments during pull request reviews and help maintain a cleaner codebase from the start.

Please, share your cool findings with me!

Source