dianakuklova

AI Agents MCP Workshop — Testing skill for Claude Code

Testing community

Learn agentic coding hands-on with Claude Code: wire up MCP servers (GitHub, Playwright), build custom agents and skills for multi-step tasks, and turn your AI assistant into a real collaborator — not.

How to install AI Agents MCP Workshop

This entry records only its repository, not the path inside it, so there is no exact command to give. Open dianakuklova/ai-agents-mcp-workshop and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AI Agents MCP Workshop does

Learn agentic coding hands-on with Claude Code: wire up MCP servers (GitHub, Playwright), build custom agents and skills for multi-step tasks, and turn your AI assistant into a real collaborator — not just autocomplete.

Alternatives in Testing

  • Ext Apps — Official repo for spec & SDK of MCP Apps protocol - standard for UIs embedded AI chatbots, served by MCP serve 1.9k ★
  • Specweave — Spec-first AI development: describe a feature → AI creates spec + plan + tasks, builds autonomously, syncs to 158 ★
  • Workflows — Catalog of multi-agent workflows for development, testing, and business tasks 127 ★

README

AI Agents, MCP & Real Work — Hands-on Assignment

How will coding look in the future? Today you'll build it. By the end of this session your AI coding assistant won't just *suggest* code — it will read issues, open pull requests, drive a browser like a real user, and refuse to let you start work that isn't properly defined. Not a shortcut. A teammate.

We use **Claude Code** as our primary tool — Anthropic's CLI that brings the full agent loop into your terminal. The same MCP servers and agentic patterns transfer to GitHub Copilot, Cursor, and any other agent-capable tool you already use. Bring your laptop. We build something real.


Before you start — prerequisites

Make sure these are ready (see the setup guide if anything is missing):

  • Node.js 18+ installed (node --version)
  • Claude Code installed and authenticated (claude --version, then claude)
  • A clean test GitHub repo you can push to, with 1–2 open issues
  • GitHub Personal Access Token available as GITHUB_PAT
  • Playwright browsers installed (npx playwright install)

Module 1 — Build something to work on

Before agents can do real work, they need something real to act on. Start by having Claude Code generate a tiny app — deliberately simple, but with an actual interaction a browser agent can later click through.

**Task:** In Claude Code, run this prompt:

Create a simple static web page as a single index.html file with the CSS inline in a  tag. No frameworks, no extra JavaScript beyond what's needed for the interaction below. Goal: it should serve as a demo that Playwright MCP will later open and click through.

Page content:
- An 

heading with the text "Workshop Demo". - A short paragraph describing the page. - A simple contact form with fields: - Name (input, id="name") - Email (input type email, id="email") - Message (textarea, id="message") - A "Submit" button (id="submit") - When Submit is clicked, show a confirmation message (