Selenium Ui Test Generator

Selenium UI Test Generator

Cost warning: UI tests cost 30-50 credits each (vs. 10-15 for unit tests). Always inform the user of this before generating.

Instructions

Step 1: Flow Selection

Ask the user to pick 2-3 critical flows maximum (e.g., login, form submission, checkout). Provide cost comparison so they can make an informed choice.

Step 2: Get Selectors Upfront

This is critical for avoiding wasted iterations. Ask the user to provide:

1. data-testid attributes (preferred)
2. Or element IDs/classes from browser dev tools

Without real selectors, tests will likely fail and waste credits. If the user cannot provide selectors, warn them of the higher failure risk before proceeding.

See references/react-testid-guide.md for the recommended data-testid pattern. Load selectors from assets/selectors-config.json when available.

Step 3: Generate ONE Test at a Time

Unlike unit/API skills, generate only 1 UI test per iteration due to the high cost.

Use explicit waits and data-testid selectors:

@Test
void testUserLogin() {
    driver.get("http://localhost:3000/login");

    WebDriverWait wait = new WebDriverWait(driver, Duration.ofSeconds(10));
    wait.until(ExpectedConditions.presenceOfElementLocated(
        By.cssSelector("[data-testid='login-form']")
    ));

    driver.findElement(By.cssSelector("[data-testid='username-input']"))
        .sendKeys("testuser@example.com");
    driver.findElement(By.cssSelector("[data-testid='password-input']"))
        .sendKeys("password123");
    driver.findElement(By.cssSelector("[data-testid='login-button']"))
        .click();

    wait.until(ExpectedConditions.urlContains("/dashboard"));
    assertTrue(driver.getCurrentUrl().contains("/dashboard"));
}

STOP. Ask the user to run the test before continuing.

Step 4: Validation

python scripts/validate_ui_tests.py 

Add --headed to see the browser during debugging.

Step 5: Fix or Skip (STRICT -- Max 2 Attempts)

If test fails:

• Attempt 1: Add explicit waits, use more specific selectors, handle React loading states
• Attempt 2: Try different selector strategy (XPath vs CSS), longer waits, check for overlays/modals
• After 2 failures: STOP. UI selector issues rarely resolve with more guessing.

Report credits used and credits saved by stopping.

Step 6: Page Object (Optional)

If the user has multiple tests for the same page, offer to create a Page Object class. This costs ~20 credits upfront but saves ~10 credits per additional test on that page.

Step 7: Summary Report

Flows Tested: N
Tests Generated: N
Tests Skipped: N (with reasons)
Credits Used: N
Credits Saved by Skipping: N

Measuring Credit Cost for Your UI Flows

Credit costs vary by Kiro account, model, and flow complexity. UI tests are the most expensive category because they generate more code (setup, waits, selectors, teardown) and fail more often (leading to retries that cost additional credits).

1. Establish your baseline (one-time)

Run the built-in benchmark with a simple login flow:

python track_credits.py benchmark ui-simple

This measures the cheapest possible UI test on your account (e.g., 0.21 credits/test).

2. Why UI tests cost more than unit or API tests

• The AI generates more code: driver setup, Chrome options, explicit waits, selector

lookups, teardown

• The AI needs more context: your selector config, page structure, wait strategies
• UI tests fail more often, and each retry costs additional credits

3. Estimate before running on your flow

More complex flows cost more:

| What makes UI tests cost more | How it's scored | |------------------------------|----------------| | More page interactions (clicks, fills) | +1 for every 3 steps | | Dynamic elements (React state, conditional rendering) | +2 for each dynamic element | | Multi-page navigation | +2 for each page transition | | More waits needed | +1 for every 3 waits |

python track_credits.py estimate \
  --type ui \
  --source-lines 50 \
  --branches 2 \
  --dynamic-selectors 1 \
  --tests 1

4. Measure actual cost after running

# Note credit balance before and after, then record:
python track_credits.py add "LoginFlow-1test"   "1 Selenium test, 4 selectors"

Your measurements improve future estimates automatically. See ../../../skills-workshop/credit-estimation/CREDIT-ESTIMATION-GUIDE.md for the full guide.

References

• references/examples/LoginFlowTest.java - Complete Selenium example
• references/react-testid-guide.md - Guide for adding data-testid to React components
• assets/selectors-config.json - Example selector configuration
• ../../../skills-workshop/credit-estimation/CREDIT-ESTIMATION-GUIDE.md - Credit measurement methodology
• Selenium WebDriver Docs
• data-testid Pattern

Source & license

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

• Author: timwukp
• Source: timwukp/agent-skills-best-practice
• License: MIT

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