Skill · QA testing
QA testing.
QA that scales with the stakes: smoke, standard, full.
Verify that a page, feature, or site is working before declaring it shipped. QA scales with the stakes across three tiers: a 2-minute smoke test after every deploy, a 10-minute standard audit for new work, and a 30-minute full release matrix before a launch. It is console-snippet driven for speed.
This is general QA, faster than the deep specialists. It catches the broken canonical, the missing alt text, and the invalid schema that visual review misses, then routes the deep WCAG and performance work to the skills built for it.
Audience: developers and QA running a post-deploy check, verifying a new page, or working through a pre-launch matrix.
The framework
Three tiers. Pick the one the stakes warrant.
QA scales with context. Run the lightest tier that matches the risk, and stop deeper testing if smoke fails.
- 01Smoke (2 minutes, after every deploy): critical signals only. Title present and 30 to 60 characters, canonical on production, exactly one H1, zero missing alts, zero broken images, expected schema types.
- 02Standard (10 minutes, new page or feature): the smoke checks plus meta description, Open Graph and Twitter tags, descriptive H2s, noopener on external links, a declared page language, and a favicon.
- 03Full (30-plus minutes, major release or pre-launch): everything above plus Lighthouse accessibility and performance, mobile and cross-browser, forms, links, sitemap, robots.txt, security headers, HTTPS, 404 handling, schema validation, and analytics.
Specific QA checks
Five targeted checks beyond the tier sweep.
Each runs as a console snippet so the result is data, not an eyeball judgment. The skill returns the output; the judgment is reading it.
01
Image audit
Every image with its source, alt text, natural dimensions, and load state, plus counts of broken and alt-less images.
02
Heading hierarchy
The full heading outline with a warning on any skipped level (an H2 jumping to an H4).
03
Contrast spot-check
WCAG contrast ratios for text-and-background pairs, flagged as pass for body, pass for large only, or fail.
04
Form audit
Per form, every field with its type, name, whether it has an associated label, and whether it is required.
05
External link audit
External links opening in a new tab that are missing noopener and noreferrer attribution.
Reference files
The reference that goes alongside the SKILL.md.
references/qa-report-template.md
A markdown report template for standard and full audits.
Bridges to other skills
When to reach for a deep specialist.
QA is fast and broad. These go deep on a single dimension when a check surfaces something that needs more than a sweep.
Deep WCAG
accessibility-auditQA runs a basic accessibility pass; a full WCAG audit with a remediation plan is a deeper, dedicated job. Reach for it when accessibility itself is the goal.
Deep performance
performance-optimizationA Lighthouse score in the full tier is a signal, not an investigation. When Core Web Vitals need real work, the performance skill goes deep.
Code and debugging
code-review-webReviewing or debugging the code behind a failing check happens there. QA verifies the rendered result, not the source.
Technical SEO baseline
seo-technicalInitial crawlability and technical SEO setup is its own baseline. QA confirms the canonical, sitemap, and robots basics on an existing build.
Open source under MIT
Read the SKILL.md on GitHub.
The skill source lives in the rampstackco/claude-skills repository alongside dozens of other skills covering the full lifecycle of brand and product work. This page is a structured overview; the SKILL.md is the source. MIT licensed.
Frequently asked questions.
- What are the three QA tiers?
- Smoke (2 minutes, run after every deploy, checking critical signals only), standard (10 minutes, for a new page or feature, covering on-page basics plus accessibility and structure), and full (30-plus minutes, for a major release or pre-launch, comprehensive across every dimension). QA scales with the stakes, so pick the tier that matches the context rather than running the heaviest one every time.
- What does a smoke test check?
- The 2-minute 'did the deploy break anything obvious' check, run as a console snippet: the title exists and is 30 to 60 characters, the canonical points at the production domain (never a staging or preview URL), there is exactly one H1, zero images are missing alt text, zero images are broken, and the schema includes the expected types for the page. If any of these fail, fix the smoke issue before any deeper testing.
- Why run a console snippet instead of eyeballing the page?
- Because visual-only QA misses the things that do not show up to the eye: missing alt text, broken or invalid schema, a missing canonical, or a canonical still pointing at staging. The console snippet surfaces all of that as data. Two related failures are running the snippet but not reading the output (the snippet is the tool; the judgment is reading what it returns) and testing in a single browser, since mobile Safari and Chrome differ enough to surprise you.
- What does the full release matrix cover?
- The smoke and standard checks plus a Lighthouse accessibility score above 90, a performance score above 80 with LCP under 2.5 seconds and CLS under 0.1, mobile responsiveness at 375, 768, 1024, and 1440 pixels, cross-browser testing in Chrome, Safari, and Firefox, form submission and validation, internal and external link checks, the sitemap and robots.txt, security headers (HSTS, X-Frame-Options, X-Content-Type-Options), HTTPS with no mixed content, correct 404 handling, schema validation, and analytics events firing on key actions.
- What happens to a failed QA check?
- It produces a fix before ship or a known-issue ticket. A failed check that ships unfixed is process theater: pass-fail with no remediation defeats the purpose of running QA at all. The most common version of this is skipping the smoke test on a 'small' deploy that looked safe, since a large share of broken-production incidents start exactly there.