Gen AI for QA: Playwright, Copilot & Claude Code

Section overview of AI-powered testing for QA engineers: how AI assists testing work, available AI tools, LLM fundamentals, and AI limitations.

AI-assisted manual testing workflow: test case and edge case generation, code fixes and explanations, test data generation, brainstorming test strategies, and data extraction from specifications, with example QA prompts.

Overview of AI tools for QA: chatbots (ChatGPT, Claude, Gemini), AI code assistants (GitHub Copilot, Claude Code, Cursor, Amazon Q, Cline, Windsurf), and specialized tools like Notebook LM, n8n, Playwright MCP, Chrome DevTools MCP, TestRigor, and TestSprite.

LLM fundamentals for testers: how large language models predict tokens, context windows, the OpenAI tokenizer, prompts versus context, custom instructions, AI agents, and agentic AI reasoning and tool use.

AI limitations in QA and test automation: token cost, hallucinations, non-deterministic outputs, verbose answers, false confidence, and privacy concerns, with mitigation strategies for each.

Prompt engineering introduction for QA: designing and refining prompts to produce accurate, relevant AI outputs using Claude, ChatGPT, or Gemini chat clients for testing tasks.

Prompt engineering principles for software testing: clarity and specificity, demonstrated with before-and-after examples generating login page test cases and bug reports in Claude.

Prompt constraints for reliable AI testing results: role assignment, scope, format, quality, and exclusion constraints, applied to a QA test-case generation prompt example.

Context engineering for QA prompts: using Playwright documentation, pasted HTML page source, ticket templates, log files, and IDE workspace files as AI context in Claude and GitHub Copilot.

Prompt patterns for QA and manual testing: zero-shot, few-shot, chain-of-thought, and iterative prompting patterns, demonstrated with bug report and ticket-writing examples in Claude.

Common prompt engineering mistakes in software testing: vague prompts, overloaded multi-part tasks, missing constraints, and insufficient context, with fixes for each.

Using Claude Code as a learning tool for QA: multiple-choice quiz generation for functional testing concepts and Selenium documentation-based quizzes.

Voice-dictated prompts in Claude for faster test automation: generating a JSON test-data array for a registration form via speech-to-text input.

Course checkpoint recapping AI-assisted QA fundamentals and prompt engineering, transitioning into hands-on AI-assisted coding sections with GitHub Copilot and Claude Code using a Playwright test suite.

Introduction to GitHub Copilot for Playwright test automation using the Awesome Pizza demo Node.js application, covering project setup and prerequisites for AI-assisted coding.

GitHub Copilot installation and configuration in Visual Studio Code: GitHub account sign-in, Agent, Ask, and Plan modes, model selection (Claude Opus), and the Web Fetch tool.

GitHub Copilot installation and setup for IntelliJ IDEA: plugin installation, GitHub device-code sign-in, inline code suggestions, and the GitHub Copilot chat panel.

Initializing a Playwright test automation project with GitHub Copilot: the npm init playwright command, TypeScript configuration, and test folder setup for the Awesome Pizza application.

Generating an end-to-end testing strategy with GitHub Copilot's fetch tool, saving the plan to a markdown file, and implementing menu.spec.ts Playwright tests iteratively.

Reviewing and fixing GitHub Copilot-generated Playwright tests: brutal-honesty code review, fixing failing assertions and locator errors, and removing hard-coded menu data for dynamic DOM checks.

Completing the Playwright test suite with GitHub Copilot: implementing order-placement spec.ts tests using the fetch tool and curl-based application discovery.

Playwright MCP server installation in VS Code for GitHub Copilot: natural-language browser interaction, saving actions as Playwright tests, and generating order-lookup spec.ts.

GitHub Copilot code assistance features: inline shadow-text suggestions, add-selection-to-chat, inline chat, and explain, review, and optimize code actions inside VS Code.

Implementing the Page Object Model pattern in Playwright with GitHub Copilot, following official Playwright documentation to refactor menu.spec.ts into a page.ts class.

Introduction to Claude Code as an agentic coding tool for test automation, reading codebases, editing files, and running commands on the Awesome Pizza Playwright project.

Claude Code installation via terminal command and VS Code extension setup, requiring a paid Claude subscription, with login authentication and dark or light theme configuration.

Basic Claude Code usage: starting sessions via terminal or VS Code extension, slash commands, and workspace file context for code explanation prompts.

Initializing a Playwright project with Claude Code via a terminal-integrated npm init playwright command, plus git add, commit, and push automation.

Generating a testing strategy with Claude Code: fetch requests to discover application functionality, saving a test plan and file plan to markdown, and iterative session refinement.

Implementing Playwright tests with Claude Code from a saved test plan: menu.spec.ts generation with Page Object Model, then removing hard-coded menu item assertions for dynamic DOM checks.

Claude Code memory and the CLAUDE.md file: auto memory versus persistent instructions, the /init command, project-level and user-level configuration, and adding TypeScript coding standards.

Completing the Playwright test suite with Claude Code from a saved test plan, generating seven remaining spec files and fixing failing notification.spec.ts assertions.

Claude Code tips: Sonnet, Opus, and Haiku model selection for task complexity, running multiple parallel sessions, and the /insights usage-statistics command.

Section overview of AI-assisted API testing: generating API collections, transforming data formats, and documenting applications with Swagger using an AI code assistant.

Generating Swagger UI API documentation with Claude Code from backend source code, producing a dependency-free swagger.json file and an API endpoint overview for the Awesome Pizza application.

Generating a Bruno API request collection with Claude Code, importing it into the VS Code Bruno extension, and debugging a broken create-order request using AI-suggested fixes.

Section overview of agentic AI and MCP servers for QA: Model Context Protocol fundamentals, Chrome DevTools MCP, and Playwright MCP for test automation.

Agentic AI versus non-agentic chat workflows for software testing: tool use, autonomous actions, and the built-in tool list inside GitHub Copilot.

Model Context Protocol (MCP) explained as a standardized interface connecting AI applications like ChatGPT, Claude Code, and GitHub Copilot to external tools, demonstrated with a custom pizza-ordering MCP server.

Finding MCP servers via the Awesome MCP Servers GitHub repository, Claude Code documentation, and the VS Code extensions marketplace, plus evaluating server safety using GitHub stars and official publishers like Microsoft's Playwright MCP.

MCP server runtime requirements: Node.js and npx for TypeScript-based servers like Playwright MCP, and Python with uv/uvx for Python-based servers like the AWS documentation MCP.

Installing the Airbnb MCP server in Claude Desktop by editing the claude_desktop_config.json file, validating JSON syntax, and testing MCP-based Airbnb search queries.

Installing the Airbnb MCP server in VS Code via a project-level .vscode/mcp.json file and a user-level global MCP configuration for GitHub Copilot.

Chrome DevTools MCP for browser automation using Puppeteer: natural-language navigation and clicks, network request and console error inspection, and Lighthouse performance audits via GitHub Copilot.

Playwright MCP installation in VS Code: natural-language browser interaction and generating Playwright test code directly from a recorded browser workflow with GitHub Copilot.

AWS Documentation MCP server setup using uvx, enabling AI-assisted search of official AWS documentation, including AWS Lambda runtime queries, inside GitHub Copilot.

Section overview of AI agent skills for QA engineers: building a Hello World skill walkthrough, then reviewing QA-relevant skill examples for test automation workflows.

AI skills defined as reusable markdown instruction files with optional scripts, loaded into context and invoked by name and description matching in Claude Code and GitHub Copilot.

Building a Hello World skill from scratch: SKILL.md structure with name and description fields, referencing a template.md response file, and executing a Node.js script to process user input.

QA-specific AI skills: a test-from-ticket skill generating structured test plans from Jira tickets, and a count-tests skill using a local script for token-efficient test file counting.

Skill creation workflow for QA: identifying workflows with clear input and output, generating SKILL.md files with AI assistance, and iterating to refine the skill.

Finding AI skills via the ClaudeSkills GitHub repository, ClaudeSkills.info, and skills.sh, including npx-based skill installation and a brainstorming skill example.

Section overview of Playwright's AI features: Playwright MCP, Playwright CLI, Playwright skills, and the Planner, Generator, and Healer test agents.

Playwright project setup for the Awesome Pizza application: npm init playwright configuration and a tsconfig.json fix for Node.js types on TypeScript.

Playwright AI accessibility-tree snapshots stored in .playwright-mcp and .playwright-cli folders, containing ARIA roles, stable IDs, and text content for token-efficient page representation.

Installing Playwright's Planner, Generator, and Healer test agents via the init agents CLI command, configuring the VS Code or Claude Code loop, and customizing agent instructions to avoid snapshot testing.

Running the Playwright Test Planner, Generator, and Healer agents end-to-end: generating a markdown test plan, implementing menu-display tests, and auto-fixing broken locators and assertions.

Playwright CLI installation via global npm package, skill installation for GitHub Copilot or Claude Code, and token-efficient skill-based commands versus MCP calls.

Playwright CLI workflow: generating a test plan, implementing tests from a test case, and healing a broken test via natural-language prompts routed through bash commands.

Token usage comparison between Playwright MCP and Playwright CLI in Claude Code using the /context command, run on the same browser theme-toggle task with the Haiku model.