Comprehensive, repeatable QA for any web application — driven by Claude Code.
It's a Claude Code plugin from Family IT Guy.
A 7-phase QA workflow that maps your app, enumerates a full test matrix, then executes it across a local → preview → prod pipeline. It covers functional, visual, performance, accessibility, security, cross-browser, failure-mode, and edge-case testing, and writes findings to a remediation-ready report.
The whole skill is built on one discipline: separate analysis from remediation. It enumerates every test, presents the full list, and lets you set scope and priority. Findings are reported as facts with evidence — no "low priority" / "not blocking" editorial — so the human makes the call.
- Measured, not vibes — every performance claim is a number (TTFB, FCP, LCP), taken uncached via CDP.
- Three engines, every pass — Chrome, Firefox, and WebKit catch three classes of bug.
- Mobile first — smallest viewport (375px) before desktop, where most users and most layout bugs are.
- Evidence-backed findings — screenshots, metrics, and
file:linereferences, each with a testable hypothesis and a verification method. - Environment progression — local → preview → real device → prod, never skipping a level, never deploying to prod without explicit approval.
Via the Family IT Guy marketplace (recommended):
/plugin marketplace add Family-IT-Guy/plugins
/plugin install qa-webapp@familyitguy
Or copy the skill directly:
git clone https://github.com/Family-IT-Guy/qa-webapp.git
mkdir -p ~/.claude/skills
cp -r qa-webapp/skills/qa-webapp ~/.claude/skills/This skill has no credentials of its own — it drives QA against whatever app you point it at, using your own test inbox and the app's own admin APIs. It does need a couple of tools:
playwright-cli(required) — all browser automation runs through it. Install the engines it drives:playwright-cli install-browser --browser=firefox playwright-cli install-browser --browser=webkit # Chrome is usually pre-installed- Lighthouse (
npx lighthouse, optional) — performance / accessibility / SEO scores. The skill falls back to the Playwright Performance API where Lighthouse can't authenticate. - Node 18+ — for Lighthouse and the optional Stripe-checkout E2E driver.
Once installed, ask Claude Code to run it:
"QA this site" · "run QA" · "is this production ready?" · "run an accessibility check"
Or invoke explicitly: /qa-webapp
Use plus-aliases on your own inbox (you+qa…@yourdomain.com) so QA emails
reach you and stay searchable. For fast-path authenticated testing, the skill
calls the target app's own admin API (e.g. a Supabase/Clerk/Auth0 admin
token) to mint sessions — supply that key via env var when you run a pass.
Nothing is baked into the skill.
skills/qa-webapp/references/stripe-checkout-e2e.md covers driving a real
Stripe Checkout flow past PerimeterX / HUMAN bot protection (which otherwise
silently rejects coupons in automation). It needs Patchright
(npm install patchright) plus a small per-project driver script — there is no
shared binary, because selectors, coupons, and URLs are project-specific. Only
relevant if your QA target uses a Stripe-hosted or embedded checkout.
qa-webapp/
├── .claude-plugin/plugin.json
├── README.md # this file
└── skills/qa-webapp/
├── SKILL.md # core instructions + 7-phase workflow
└── references/
├── checklists.md # every test item, per category
├── methodology.md # step-by-step execution procedures
├── benchmarks.md # pass/fail thresholds + measurement
└── stripe-checkout-e2e.md # real Stripe Checkout E2E (optional)
MIT