How to Use Selenium Skill With Claude Code [2026]

What Is a Selenium Skill

A Selenium Skill is a SKILL.md file that loads Selenium-specific rules into AI coding agents like Claude Code, Cursor, and GitHub Copilot to enforce stable locators, explicit waits, and Page Object structure in every generated test.

The skill defines when to use each API, which patterns to avoid, and what keeps a test suite maintainable. The format follows the open Agent Skills Standard. One file works in Claude Code, Cursor, GitHub Copilot, Gemini CLI, and 16+ other tools.

How Does a Selenium Skill Fix Brittle Test Code

The Selenium Skill replaces brittle CSS selectors with @FindBy locators inside Page Objects and swaps Thread.sleep() for WebDriverWait with ExpectedConditions, so generated tests survive UI changes across sprints.

Without the skill, Claude Code generates this:

driver.findElement(By.cssSelector(".btn-primary-blue")).click();
Thread.sleep(3000);

The CSS selector .btn-primary-blue breaks when a designer renames the class. Thread.sleep(3000) wastes three seconds whether the element is ready or not.

With the skill, the same prompt produces a Page Object with stable locators:

// LoginPage.java
@FindBy(id = "submit")
private WebElement submitButton;

public void clickSubmit() {
    new WebDriverWait(driver, Duration.ofSeconds(10))
        .until(ExpectedConditions.elementToBeClickable(submitButton))
        .click();
}

The skill applies these rules to the generated code:

The @FindBy(id = "submit") locator stays stable when CSS classes change. Across a full suite, this keeps tests passing through UI updates.

For a deeper look at synchronization strategies, see this guide on Selenium waits.

What Is Inside the Selenium Skill Repository

The Selenium Skill repository contains a SKILL.md entry point that defines locator and wait rules, and a reference folder with a full implementation playbook, advanced patterns for Shadow DOM and AJAX, and cloud integration config.

The selenium-skill is one of 46 skills in the TestMu AI agent-skills repository. It sits alongside Playwright, Cypress, Appium, and WebdriverIO.

Here is the folder layout:

selenium-skill/
├── SKILL.md                    # Entry point — locator logic, wait rules, framework selection
└── reference/
    ├── playbook.md             # Java/Python/JS setup, POM templates, CI configs, debug table
    ├── advanced-patterns.md    # Shadow DOM, AJAX handling, custom waits, parallel execution
    └── cloud-integration.md    # TestMu AI capabilities, credentials, tunnel, dashboard

Claude Code reads SKILL.md first. It stays under 500 lines to avoid bloating the context window.

When the agent needs deeper guidance, it pulls from the reference files on demand.

How to Install the Selenium Skill in Claude Code

Clone the TestMu AI agent-skills repository into your project's .claude/skills/ directory using git clone. Claude Code detects the skill automatically on the next session start with no extra configuration.

git clone https://github.com/LambdaTest/agent-skills.git .claude/skills/agent-skills

That gives Claude Code every skill in the repository. If you only want Selenium:

git clone https://github.com/LambdaTest/agent-skills.git
cp -r agent-skills/selenium-skill .claude/skills/

The same approach works for other tools. Replace .claude/skills/ with the target path:

• Cursor: .cursor/skills/agent-skills
• GitHub Copilot: .github/skills/agent-skills
• Gemini CLI: .gemini/skills/agent-skills

Once the folder is in place, every new session picks up the skill automatically.

How to Run Tests Using Selenium Skills in Claude Code

Open Claude Code and give it a test target with the language and framework you want. The agent reads the Selenium Skill and generates the full project for you.

I entered the following prompt:

Write a test in Selenium Java that visits https://ecommerce-playground.lambdatest.io/ and clicks on the Blog link in the header.

Claude Code picked up the Selenium Skill and generated four files:

• pom.xml: Selenium 4, TestNG, WebDriverManager dependencies
• pages/HomePage.java: Page Object with @FindBy and PageFactory
• tests/BlogNavigationTest.java: Test class with @Test and @BeforeMethod annotations
• testng.xml: Suite configuration for execution

Next, I passed another prompt to run the test:

Run the test

I did not specify where to run it. By default, the Selenium Skill executes tests on TestMu AI, a cloud testing platform that provides an online Selenium Grid with 3000+ browser and OS combinations. So the test ran on the cloud without any extra setup on my end.

Note: For cloud execution, you need a TestMu AI account to set your credentials in the environment variables. You can find them at the bottom of the TestMu AI Home dashboard under Credentials.

# macOS/Linux
export LT_USERNAME="YOUR_TESTMUAI_USERNAME"
export LT_ACCESS_KEY="YOUR_TESTMUAI_ACCESS_KEY"

# Windows (PowerShell)
$env:LT_USERNAME="YOUR_TESTMUAI_USERNAME"
$env:LT_ACCESS_KEY="YOUR_TESTMUAI_ACCESS_KEY"

# Windows (Command Prompt)
set LT_USERNAME="YOUR_TESTMUAI_USERNAME"
set LT_ACCESS_KEY="YOUR_TESTMUAI_ACCESS_KEY"

The TestMu AI Web Automation dashboard showed the test passed. The report included execution time, command steps, and a video replay confirming the Blog link click worked as expected.

If a test fails, you paste the stack trace back into Claude Code.

For more details, check out this guide to run Selenium tests using Agent Skills.

How to Author Selenium Tests at Scale

A Selenium Skill improves test quality, but when your team cannot author tests fast enough, coverage gaps grow every sprint.

Generative AI testing tools like TestMu AI's KaneAI solve this from the other side. Instead of an engineer prompting Claude Code test by test, KaneAI takes natural language prompts describing the user flow and generates the full Selenium test, ready to export.

Here is what it brings to Selenium teams specifically:

• Natural Language to Selenium Code: Describe a flow in plain English and KaneAI converts it into executable Selenium test steps.
• Direct Selenium Export: Export tests as Selenium Java, Python, or C# that plug into your existing Maven or Gradle project.
• 2-Way Editing: Switch between the plain-language view and the Selenium code view while keeping both in sync.
• Self-Healing Locators: KaneAI resolves broken @FindBy selectors automatically when UI elements change.
• Cross-Browser Execution: Run tests across Chrome, Firefox, Safari, and Edge on the TestMu AI grid.
• Assisted Debugging: Analyzes failing Selenium commands in real time and surfaces root cause with a fix.
• Multi-Input Test Generation: Feed a Jira ticket, PRD, screenshot, or spreadsheet to generate Selenium tests automatically.

Conclusion

The Selenium Skill gives Claude Code the rules your team already follows. Stable locators in Page Objects, explicit waits, and credentials in environment variables.

Install the skill, point Claude Code at a page, and the output is a structured test project ready to run.

The TestMu AI agent-skills ships the Selenium Skill alongside 45 others covering Playwright, Cypress, Appium, and more.