TaskMarket: Building an Taks Marketplace with GitHub Copilot CLI

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I built TaskMarket, a MERN web app where people can post paid needs (tasks) and others can submit offers to fulfill them.

A key rule in this project is that payment happens offline, not on the platform.

Core flow

• Requesters create tasks with details and budget.
• Fulfillers browse tasks and submit offers.
• Requesters accept an offer and coordinate through chat.
• Task closure is confirmed only after completion/offline payment confirmation.
• Reviews are supported after task closure.

Why this project matters to me

I wanted to build a practical marketplace-style app that focuses on trust, workflow integrity, and realistic constraints (like offline payment), while using Copilot CLI heavily in an end-to-end build process.

Demo

Hosted URL: https://task-market-client.vercel.app/

Demo login users

1. User: [e-mail] / 12345678
2. User: [e-mail] / 12345678
3. Admin: [e-mail] / Brillant123!

My Experience with GitHub Copilot CLI

I used GitHub Copilot CLI with gpt-5.3-codex (medium and high) throughout development.

My prompt journey

1. Opened VS Code with an empty workspace.
2. Launched Copilot CLI in terminal to connect with VS Code.
3. Switched to Plan mode.
4. Prompt: “A website where users can post their needs and are willing to pay for it, others can fulfil their needs and get paid but payment will be done offline not on website.”
5. Prompt: “install the necessary copilot/AI agent skills required in this project by looking them online”
   • Copilot CLI generated taskmarket-builder.agent.md, copilot-instructions.md, and AGENTS.md.
6. Prompt: “Look information online and install the AI agents skills for this project, I believe frontend-dev, backend-dev, react-skill will be necessary”
   • It created .github/agents instead of skills.
7. Prompt: “install skills not GitHub Copilot agents”
   • It created skills itself instead of downloading popular presets.
8. Prompt: “Don't create skills yourself, download them as it is from internet, research online what skill are available on internet using subagents”
   • I realized I was in Plan mode, deleted created skills, switched to Agent mode, and ran /new.
9. Prompt: “Research online using subagents and download the Claude/Copilot/AI agents skill which would be good while developing the @plan.md like frontend/backend/react skills. Try to install them by running cli commands instead of fetching and storing them in files, refer: https://github.com/skillcreatorai/Ai-Agent-Skills”
   • Copilot CLI sessions crashed repeatedly. I reported it here: https://github.com/github/copilot-cli/issues/1468
10. I manually installed skills by downloading from the internet into .github/skills (frontend-design and backend-development).
11. Deleted agents created by Copilot CLI.
12. New session prompt: “Create tasks for the @plan.md”
    • Copilot updated @plan.md with tasks.
13. Prompt: “start implementing the workplan”
14. Prompt: “create tasks for @plan.md”
15. Prompt: “implement workplan”
16. Prompt: “use frontend skill to improve the frontend design”
17. Prompt: “Generate fake data for website”
18. Prompt: “continue workplan”
19. Prompt: “continue workplan”

Impact on my development

• Copilot CLI helped me quickly scaffold, plan, and iterate across backend + frontend.
• The plan-driven workflow made implementation more structured.
• There were friction points (especially around skill installation and session stability), but overall it accelerated development significantly once the setup stabilized.