Webapp Testing

Web Application Testing

Test local web applications using browser automation.

Tool Selection

| Use Case | Tool | Why | |----------|------|-----| | AI agent browser interaction | agent-browser | AI-optimized refs, low token cost | | Automated test suites (CI/CD) | Playwright | Rich assertions, parallelization | | PDF generation from HTML | Puppeteer | Specialized PDF APIs | | Visual regression testing | Playwright | Screenshot comparison |

Default: Use agent-browser for interactive testing, Playwright for test suites.

Decision Tree

1. Is this AI-driven exploration or automated test suite?

• AI exploration: Use agent-browser (refs system, AI-optimized)
• Automated tests: Use Playwright (assertions, CI/CD)

1. Is target static HTML or dynamic webapp?

• Static: Direct page load and inspect
• Dynamic: Use with_server.py helper for server lifecycle

1. Do you need visual or programmatic testing?

• Visual: Screenshots and visual comparison
• Programmatic: DOM queries and assertions

---

Agent-Browser (AI Agent Testing)

Install globally: bun add -g agent-browser

Quick Start

# Navigate and get AI-optimized snapshot
agent-browser navigate "http://localhost:3000"
agent-browser snapshot

# Interact using refs from snapshot
agent-browser click @e3
agent-browser fill @e5 "test@example.com"
agent-browser snapshot  # Verify changes

The Refs Advantage

After snapshot, each element gets a unique ref (@e1, @e2):

• Deterministic: No CSS selector guessing
• Fast: No DOM re-querying
• AI-friendly: LLMs parse refs reliably

Common Patterns

# Login flow
agent-browser navigate "http://localhost:3000/login"
agent-browser snapshot
agent-browser fill @e2 "user@example.com"
agent-browser fill @e3 "password123"
agent-browser click @e4
agent-browser wait-for "Dashboard"
agent-browser snapshot

# Form submission
agent-browser navigate "http://localhost:3000/contact"
agent-browser snapshot
agent-browser fill @e1 "John Doe"
agent-browser fill @e2 "john@example.com"
agent-browser type @e3 "Hello, this is my message"
agent-browser click @e4  # Submit button
agent-browser wait-for "Thank you"

See ~/.claude/context/agent-browser.md for full command reference.

---

Playwright (Automated Test Suites)

Setup

bun add -D playwright @playwright/test
bunx playwright install

Basic Test Structure

import { test, expect } from '@playwright/test';

test('homepage loads correctly', async ({ page }) => {
  await page.goto('http://localhost:3000');

  // Wait for dynamic content
  await page.waitForLoadState('networkidle');

  // Check title
  await expect(page).toHaveTitle(/My App/);

  // Check element exists
  await expect(page.locator('h1')).toContainText('Welcome');
});

Critical Practice

Always wait for network idle on dynamic apps:

await page.waitForLoadState('networkidle');

This prevents testing incomplete page states caused by ongoing JavaScript execution.

Element Selection

// By text
await page.getByText('Submit').click();

// By role
await page.getByRole('button', { name: 'Submit' }).click();

// By test ID (recommended)
await page.getByTestId('submit-button').click();

// By CSS selector
await page.locator('.submit-btn').click();

// By placeholder
await page.getByPlaceholder('Enter email').fill('test@example.com');

Screenshots

// Full page
await page.screenshot({ path: 'screenshot.png', fullPage: true });

// Specific element
await page.locator('.header').screenshot({ path: 'header.png' });

// Hide elements
await page.screenshot({
  path: 'clean.png',
  mask: [page.locator('.ads')]
});

Form Interaction

// Fill form
await page.getByLabel('Email').fill('user@example.com');
await page.getByLabel('Password').fill('secret123');

// Select dropdown
await page.getByLabel('Country').selectOption('US');

// Checkbox
await page.getByLabel('Remember me').check();

// Submit
await page.getByRole('button', { name: 'Sign In' }).click();

API Mocking

await page.route('**/api/users', async (route) => {
  await route.fulfill({
    status: 200,
    contentType: 'application/json',
    body: JSON.stringify([{ id: 1, name: 'Test User' }])
  });
});

Console Logging

page.on('console', msg => {
  console.log(`Browser console [${msg.type()}]: ${msg.text()}`);
});

page.on('pageerror', err => {
  console.error('Page error:', err.message);
});

Testing Dynamic Content

// Wait for element to appear
await page.waitForSelector('.loading', { state: 'hidden' });
await page.waitForSelector('.content', { state: 'visible' });

// Wait for specific text
await page.waitForFunction(() =>
  document.body.innerText.includes('Data loaded')
);

// Wait for network request
const responsePromise = page.waitForResponse('**/api/data');
await page.click('#load-button');
const response = await responsePromise;

Workflow Pattern

1. Reconnaissance: Capture screenshots, inspect DOM
2. Identify: Find selectors from rendered state
3. Execute: Automate using discovered selectors
4. Verify: Assert expected outcomes

// Reconnaissance
await page.screenshot({ path: 'before.png' });
console.log(await page.content());

// Find and interact
const buttons = await page.locator('button').all();
for (const btn of buttons) {
  console.log(await btn.textContent());
}

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Lucface
• Source: Lucface/claude-skills
• License: MIT

Install and usage instructions live in the source repository linked above.