flaky-test-analyzer

name	Flaky Test Analyzer
description	Skill phân tích và khắc phục các automation test không ổn định (flaky tests), xác định root cause và đề xuất fix.

Flaky Test Analyzer

Purpose: Identify and resolve unstable automation tests.

---

When to Use

Use this skill when:

• A test passes and fails intermittently
• Test results are inconsistent across runs
• CI/CD pipeline has unreliable test results

---

Responsibilities

Detect flaky tests caused by:

• Unstable locators (dynamic classes, positional xpath)
• Timing issues (race conditions, slow page loads)
• Incorrect waits (hard sleep instead of smart waits)
• Environment dependency (data not cleaned up, external service down)
• Test data conflicts (shared data between parallel tests)

---

Analysis Workflow

1. Detect — Identify the failing test and reproduce the failure
2. Inspect — Read error logs, stack traces, and screenshots
3. Classify — Categorize the root cause (locator / timing / data / environment)
4. Fix — Apply the appropriate fix strategy
5. Verify — Re-run test multiple times to confirm stability

---

Common Flaky Causes & Fixes

Unstable Locator

Problem:

//div[3]/button
.css-1n2xyz-btn

Fix: Replace with stable locator following priority in .claude/rules/locator_strategy.md:

• id, data-testid, name, css selector (stable), xpath (relative)

---

Timing Issues

Problem:

Thread.sleep(3000);       // Hard sleep — BAD
page.waitForTimeout(2000); // Fixed delay — BAD

Fix: Use smart waits as defined in .claude/rules/selenium_rules.md and .claude/rules/playwright_rules.md:

// Selenium
WebDriverWait wait = new WebDriverWait(driver, Duration.ofSeconds(10));
wait.until(ExpectedConditions.visibilityOfElementLocated(By.id("result")));

// Playwright
await expect(page.getByRole('button', { name: 'Submit' })).toBeVisible();

---

Test Data Conflicts

Problem: Tests share mutable data → parallel runs conflict.

Fix: Use unique, traceable random data:

<testName>_<timestamp>@test.com

---

Stability Checklist

After fixing a flaky test, verify:

• Locator is unique and stable across reloads
• No hard sleep or fixed delays
• Test data is unique and deterministic
• Test is independent (no dependency on other tests)
• Test passes 5+ consecutive runs

---

Rules References

The agent MUST follow these rules when analyzing flaky tests:

• .claude/rules/locator_strategy.md — Locator stability rules
• .claude/rules/automation_rules.md — General automation best practices
• .claude/rules/selenium_rules.md — Selenium wait strategy
• .claude/rules/playwright_rules.md — Playwright auto-waiting