What does this skill do?

QA Skill Test Automation runs visual, interaction, and accessibility tests in the browser after a deployment, simulating real user actions to detect errors before they reach production. It covers smoke tests, visual regression testing across multiple resolutions, validation of critical workflows, and WCAG 2.2 AA compliance audits, generating a structured report with a clear verdict.

Post-Deployment Smoke Test
Navigate to the target URL, check for console errors, failed requests, and Core Web Vitals (LCP, CLS, INP).
Interaction Tests
Check navigation links, submit forms with valid and invalid data, and test complete authentication flows.
Multiresolution Visual Regression
Take screenshots at 375px, 768px, and 1440px, compare them with the previous version, and check the dark mode if applicable.
Accessibility Audit
Run axe-core to detect WCAG 2.2 AA violations: contrast, labels, focus order, and keyboard navigation.

Usage examples

🚀 Complete smoke test
Run a smoke test at https://staging.midominio.com after the latest deployment and check the console, network, and Core Web Vitals.
📝 Submit the contact form
Test the contact form on the staging environment by submitting it with no data and with invalid data. Verify that the correct error messages appear.
📸 Visual regression at 3 resolutions
Take screenshots of the key pages at 375px, 768px, and 1440px, and compare them to the previous visual reference.
♿ WCAG Accessibility Audit
Conduct an accessibility audit on the home page and check the contrast, labels, and keyboard navigation.

Features

Automated Smoke Test Check for console errors, failed network requests, and Core Web Vitals in a single pass.
Real-world interaction tests Simulate clicks, form submissions, and authentication flows to detect broken links and error states.
Comparative Visual Regression Take screenshots at 375px, 768px, and 1440px and compare them to a previous baseline image to detect changes.
WCAG 2.2 AA Audit Use Axe-Core to detect accessibility violations: contrast, alternative text, focus order, and keyboard navigation.
Markdown Report with Verdict Generate a structured report with a final verdict: PUBLISH, PUBLISH WITH CORRECTIONS, DO NOT PUBLISH, or INCONCLUSIVE.

Frequently asked questions

You need a browser automation MCP such as Playwright or Puppeteer set up in Claude to run the tests.
By default, it operates in read-only mode. Never perform actions that modify data (payments, deletions) in production without explicit authorization, and always use test credentials.
The skill marks the visual regression result as "INCONCLUSIVE" if there is no previous baseline image to compare it to.
The final verdict may be "PUBLISH," "PUBLISH WITH REVISIONS," "DO NOT PUBLISH," or "INCONCLUSIVE," depending on the number and severity of the problems detected.
QA Test Automation — Verify Your UI After Deployments with Claude AI

¿Prefieres escuchar el contenido? Genera la narración de audio con un clic.