Agent Native

playwright-skill

The skill executes Playwright scripts for browser automation, testing web applications, and validating functionality.

by lackeyjbCommunityNot reviewed

Community author

Agents
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode
License
MIT (permissive)
Pin
dd47a6apinned Oct 9, 2026
Context cost
about 1,800 tokenswhen the skill loads
GitHub stars
3,180on lackeyjb/playwright-skill
Scripts
2 scripts

What it does

“Complete browser automation with Playwright. Auto-detects dev servers, writes reusable test scripts, and supports screenshots, responsive checks, UX validation, login flows, link checks, and arbitrary browser automation. Use when the user wants to test a website, automate browser interactions, validate web functionality, or perform browser-based testing.”
The author's description, from SKILL.md

Generated summary from the skill's own text

The skill executes Playwright scripts for browser automation, testing web applications, and validating functionality. It detects local development servers, generates reusable test scripts, and provides helpers for screenshots, responsive testing, login flows, and arbitrary browser interactions.

When to use it

  • Test website functionality and user workflows in a browser
  • Validate responsive design across different screen sizes
  • Automate login flows and form submissions
  • Check links, content rendering, and page state

When not to

  • Testing applications requiring browsers not supported by Playwright
  • Automating interactions with applications outside web browsers
  • Verifying complex visual layouts requiring manual review

Install

These commands install from the publisher's repository and need no account.

skills CLI

npx skills add lackeyjb/playwright-skill --skill playwright-skill

Installs from the repository's default branch, which may have changed since the commit we checked.

GitHub CLI 2.90 or later

gh skill install lackeyjb/playwright-skill playwright-skill@dd47a6a023e249eb1b36e9e943eab89d0900865d

Installs the commit we checked (dd47a6a).

Copy by hand

Download the folder at dd47a6a and put it in .claude/skills/playwright-skill for Claude Code and in .agents/skills/playwright-skill for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode.

Skill folder on GitHub at dd47a6a

Copy for your agent

A prompt for any coding agent. It installs this skill from GitHub at the pinned commit, copies the files as they are, and runs nothing unless you ask.

Prompt

Install the agent skill "playwright-skill" by lackeyjb.

Source, pinned to commit dd47a6a023e249eb1b36e9e943eab89d0900865d:
https://github.com/lackeyjb/playwright-skill/tree/dd47a6a023e249eb1b36e9e943eab89d0900865d/skills/playwright-skill

1. Download that folder at exactly this commit. Do not use the default branch.
2. Copy every file unchanged, keeping the folder structure, into .claude/skills/playwright-skill/ for Claude Code and .agents/skills/playwright-skill/ for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode. Create the folders if they do not exist.
3. Do not edit the files, and do not run any script in the folder unless I ask you to.
4. When you are done, list the files you copied and where you put them.

Listing and checks: https://www.agentnative.dev/skills/lackeyjb/playwright-skill

SKILL.md at the pin

---
name: playwright-skill
description: Complete browser automation with Playwright. Auto-detects dev servers, writes reusable test scripts, and supports screenshots, responsive checks, UX validation, login flows, link checks, and arbitrary browser automation. Use when the user wants to test a website, automate browser interactions, validate web functionality, or perform browser-based testing.
license: MIT
compatibility: Requires Node.js 20+, npm, and network access on first setup to install Playwright and Chromium.
metadata:
  author: lackeyjb
  version: "5.0.0"
allowed-tools: Bash(node:*) Bash(npm:*) Read Write
---

# Playwright Browser Automation

Write and execute focused Playwright scripts for the user's request. Prefer the
skill's executor and helpers, but use the full Playwright API when needed.

## Path resolution

This skill can be installed in several locations, so resolve its directory
first. Set `SKILL_DIR` to the directory containing this SKILL.md file, then run
the commands below as written:

```bash
export SKILL_DIR=<absolute path of the directory containing this SKILL.md>
export TMP_DIR="$(node -p 'require("node:os").tmpdir()')"
```

If shell state does not persist between commands, substitute the literal paths
for `$SKILL_DIR` and `$TMP_DIR` in each command instead.

Common installation paths:

- Plugin system: `~/.claude/plugins/marketplaces/playwright-skill/skills/playwright-skill`
- Manual global: `~/.claude/skills/playwright-skill`
- Project-specific: `<project>/.claude/skills/playwright-skill`

## Workflow

1. For localhost work, detect running servers before writing a URL:

   ```bash
   node -e "require('$SKILL_DIR/lib/helpers').detectDevServers().then(s => console.log(JSON.stringify(s)))"
   ```

   Use the only result automatically. Ask which URL to use when there are
   multiple results. Ask for a URL or offer to start a server when none exist.
2. Write reusable scripts to `$TMP_DIR/playwright-test-*.js` unless the user
   asks to save them in the project. Use `PW_SCRIPT_DIR` to preserve scripts.
3. Use a visible browser by default. Use `headless: true` only when requested
   or when the environment has no display.
4. Put the target URL in a constant or environment variable.
5. Run scripts with `node "$SKILL_DIR/run.js" <script.js>`.
6. Report actions, failures, and artifact paths. Do not claim success without
   checking the resulting page.

## Setup

Run once:

```bash
cd "$SKILL_DIR" && npm run setup
```

This installs Playwright and Chromium. Use `cd "$SKILL_DIR" && npm run
install-all-browsers` when Firefox or WebKit is required.

## Minimal example

```javascript
const os = require('node:os');
const path = require('node:path');
const { chromium } = require('playwright');

const targetUrl = process.env.TARGET_URL || 'http://localhost:3000';
const artifactDir = process.env.PW_ARTIFACT_DIR || os.tmpdir();

(async () => {
  const browser = await chromium.launch({ headless: false });
  try {
    const page = await browser.newPage();
    await page.goto(targetUrl);
    console.log('Page loaded:', await page.title());
    await page.screenshot({ path: path.join(artifactDir, 'page.png'), fullPage: true });
  } finally {
    await browser.close();
  }
})();
```

Run it:

```bash
node "$SKILL_DIR/run.js" "$TMP_DIR/playwright-test-page.js"
```

For short one-off tasks, use inline execution:

```bash
node "$SKILL_DIR/run.js" -e "const browser = await chromium.launch({headless: false}); try { const page = await browser.newPage(); await page.goto('https://example.com'); console.log(await page.title()); } finally { await browser.close(); }"
```

The `-e` process exits as soon as the snippet settles, so close the browser
inside the snippet.

## Current Playwright patterns

Prefer locators that describe what a user sees, in this order:

1. `page.getByRole()` with an accessible name
2. `page.getByLabel()` for form controls
3. `page.getByText()` for visible content
4. `page.getByTestId()` when the application provides a test contract

Actions auto-wait for actionability. Use web-first assertions or a locator's
`waitFor()` instead of `waitForSelector()`, fixed sleeps, or `networkidle`.

```javascript
await page.getByLabel('Email').fill('test@example.com');
await page.getByRole('button', { name: 'Sign in' }).click();
await page.waitForURL('**/dashboard');
await page.getByRole('heading', { name: 'Dashboard' }).waitFor();
```

## Common tasks

### Responsive checks

```javascript
{
  const os = require('node:os');
  const path = require('node:path');

  const artifactDir = process.env.PW_ARTIFACT_DIR || os.tmpdir();
  const viewports = [
    { name: 'desktop', width: 1440, height: 900 },
    { name: 'mobile', width: 390, height: 844 },
  ];

  for (const viewport of viewports) {
    await page.setViewportSize(viewport);
    await page.goto(targetUrl);
    await page.screenshot({ path: path.join(artifactDir, `${viewport.name}.png`), fullPage: true });
  }
}
```

### Login flow

Use test credentials supplied by the user. Never invent or expose real
credentials. Verify both the navigation and a post-login element.

```javascript
await page.goto(`${targetUrl}/login`);
await page.getByLabel('Email').fill(process.env.TEST_EMAIL);
await page.getByLabel('Password').fill(process.env.TEST_PASSWORD);
await page.getByRole('button', { name: /sign in|log in/i }).click();
await page.waitForURL('**/dashboard');
await page.getByRole('heading', { name: /dashboard/i }).waitFor();
```

### Save scripts and artifacts

```bash
PW_SCRIPT_DIR=./playwright-tests node "$SKILL_DIR/run.js" "$TMP_DIR/playwright-test-login.js"
PW_ARTIFACT_DIR=./playwright-artifacts node "$SKILL_DIR/run.js" "$TMP_DIR/playwright-test-page.js"
```

`PW_SCRIPT_DIR` copies file-based scripts before execution and adds a timestamp
when a filename already exists. `PW_ARTIFACT_DIR` controls helper screenshot
output; the default is the operating system temporary directory.

### Connect to an existing Chrome session

Start Chrome with remote debugging enabled, then connect with Playwright:

```javascript
const browser = await chromium.connectOverCDP('http://127.0.0.1:9222');
const page = browser.contexts()[0].pages()[0];
```

This reuses cookies and extensions in that session. Do not use it for secrets
unless the user explicitly asks; a connected browser has the user's access.

## Helpers

```javascript
const helpers = require(`${process.env.PW_SKILL_DIR}/lib/helpers`);

const servers = await helpers.detectDevServers();
const browser = await helpers.launchBrowser('chromium');
const context = await helpers.createContext(browser);
const page = await context.newPage();
await helpers.handleCookieBanner(page);
await helpers.takeScreenshot(page, 'result');
```

Available helpers are `detectDevServers`, `getExtraHeadersFromEnv`,
`launchBrowser`, `createContext`, `handleCookieBanner`, and `takeScreenshot`.
Use Playwright locators and assertions directly for actions, waits, extraction,
authentication, tables, and retries.

## Configuration

- `PW_BROWSER`: `chromium`, `firefox`, or `webkit` for `launchBrowser()`.
- `PW_CHANNEL`: installed browser channel such as `chrome` or `msedge`.
- `PW_EXECUTABLE_PATH`: explicit browser executable path.
- `PW_HEADLESS`: `true` or `false`; visible mode is the default.
- `SLOW_MO`: action delay in milliseconds.
- `PW_HEADER_NAME` and `PW_HEADER_VALUE`: one extra HTTP header.
- `PW_EXTRA_HEADERS`: JSON object of extra HTTP headers.
- `PW_SCRIPT_DIR`: directory for preserving file-based scripts.
- `PW_ARTIFACT_DIR`: directory for helper-generated screenshots.

See [API_REFERENCE.md](API_REFERENCE.md) for network interception, API mocking,
authentication state, video, visual checks, device emulation, and CI patterns.

From lackeyjb/playwright-skill at dd47a6a, shown as plain text under its license: MIT (permissive). The license text is in skills/playwright-skill/SKILL.md.

View SKILL.md on GitHub at dd47a6a

Files

6 files, 35 KB in total. Each file has a sha256 in the pin's manifest; the first 12 characters are shown.

Skill folder on GitHub at dd47a6a
All Testing and QA skills
  • triaging-visual-review-runs

    Inspects PostHog Visual Review screenshot regression runs that gate pull requests, helping engineers triage diff results, identify flaky stories, and decide…

    by PostHogOfficial publisher

    Claude Code, Codex, Cursor +4 · about 8.9K tokens · pinned 6 days ago

  • webapp-testing

    This skill enables agents to test local web applications using Playwright by writing Python automation scripts.

    by SnykOfficial publisher

    Contains scripts

    Claude Code, Codex, Cursor +3 · about 890 tokens · pinned 6 days ago

  • browser-probe

    This skill probes a URL with escalating headless browser configurations to detect CDN bot protection systems like Akamai and Cloudflare, then generates a…

    by Adobe, Inc.Official publisher

    Contains scripts

    Claude Code · about 1K tokens · pinned 6 days ago

  • browser-navigation

    This skill enables agents to automate browser interactions including navigation, form filling, element clicking, screenshots, and data extraction from web…

    by FactoryOfficial publisher

    Claude Code, Codex, Cursor +3 · about 2.5K tokens · pinned 6 days ago

  • triaging-visual-review-runs

    Inspects PostHog Visual Review runs that gate PR merges with screenshot regression checks.

    by PostHogOfficial publisher

    Claude Code, Codex, Cursor +4 · about 8.9K tokens · pinned 6 days ago

  • clerk-testing

    This skill enables end-to-end testing of Clerk authentication flows using Playwright or Cypress test frameworks.

    by ClerkOfficial publisher

    Claude Code, Codex, Cursor +4 · about 480 tokens · pinned 6 days ago

The skills listed here are third-party software, collected from public GitHub repositories as a catalog of recent skills and provided as is. Agent Native does not write or control them and is not responsible for any harm, loss or damage caused by installing or using them. Labels and checks describe what we saw at one commit and are not a guarantee. Read a skill's files before you run it. Terms of use