Web QA testing skills for AI coding agents — browser flows, accessibility, visual regression, performance, SEO, i18n, bug triage, and Jira filing, driven by Playwright.
Skills 30 in the Claude listing
Accessibility audit Run an accessibility audit against a live site using a real browser (via Playwright MCP or @playwright/cli), combining automated axe-core scans with a defined set of manual checks axe can't do, and ou Auth session audit Test session and authentication lifecycle edge cases using a real browser — session expiry mid-flow, concurrent sessions across simulated devices, whether logout in one tab propagates to another, reme Browser safety Use before driving a browser for any testing task — the credential, screenshot, and untrusted-content rules every browser-driving check in this toolkit depends on. Bug analyzer Use when the user shares a bug report, stack trace, failing test, or Jira issue and wants likely root causes and related code identified — returns ranked, evidence-cited hypotheses (the file:line, the Bug creator Use when a defect needs to become a Jira ticket — takes a finding from any source (a bug-triage bundle, a drafted ticket markdown, a figma-compare/fix-verifier finding, or a defect described in conver Bug triage Reproduce a reported bug in a real browser (via the Playwright MCP server or @playwright/cli), determine how reliably it reproduces, capture a full evidence bundle (trace, screenshot, console, network Cross browser matrix Run an existing test scenario across a matrix of browser engines (Chromium, Firefox, WebKit) and viewport sizes (mobile/tablet/desktop), and normalize the results into a single pass/fail comparison ta Crud tester Test create, edit, and delete (CRUD/mutation) operations in a web app with a real browser, verifying the full data lifecycle of each entity and cleaning up via delete-as-test. E2e codegen Convert a scenario that's already been verified by flow-runner (or network-assertion / cross-browser-matrix for their scenario types) into a permanent, runnable @playwright/test spec file — with role- Email verification Verify the email half of a web flow end to end — trigger the flow in a real browser (signup, invite, password reset, order confirmation, notification), confirm the email actually arrives in a test-con Figma compare Compare a Figma design against its live web implementation using the Figma MCP server (design side) and a real browser via Playwright MCP or @playwright/cli (implementation side), producing an evidenc Fix verifier Re-verify previously-found defects against the current build using a real browser — take one or more defect IDs (or "all open defects"), locate each one's recorded minimal repro, re-run it faithfully Flow runner Drive and verify end-to-end user flows (login, signup, checkout, search, multi-step forms/wizards, etc.) in a real browser using Playwright, via either the Playwright MCP server or the @playwright/cli Form fuzzing Test form fields with unusual-but-realistic inputs (empty, boundary-length, unicode, malformed formats, real-but-uncommon characters like apostrophes in names) using a real browser, to verify the app I18n audit Check internationalization correctness across a site's supported locales using a real browser — text actually changes per locale, no leaked raw translation keys, right-to-left layouts don't visually b Jira connector Use when the user wants to read from or write to Jira Cloud — fetch an issue, run a JQL search, list projects/fields/transitions, or create issues, comments, and status transitions. Link audit Crawl a site exhaustively (deeper and wider than scenario-mapper's shallow, curated pass) using a real browser to discover every internal link and image, then verify each one's status via lightweight Network assertion Observe, assert on, and mock network/API traffic during a browser session using Playwright (route interception via @playwright/cli, the MCP server's browser_route/network tools, or raw page.route()) — Performance audit Measure Core Web Vitals (LCP, INP, CLS) and related performance signals on a site's key pages using a real browser (via Playwright MCP or @playwright/cli), by injecting the web-vitals library the same Qa onboarding Write an onboarding document for a new QA teammate joining this team's testing practice — the site's testing landscape (scope, P0 areas, cadence, environment/test data conventions, pulled from test-pl Report generator Synthesize outputs from the other testing skills in this toolkit (scenario CSVs, run reports, bug-triage writeups, audit findings, cross-browser matrices) — potentially from several sessions across a Scenario mapper Explore a live website with a real browser (via the Playwright MCP server or @playwright/cli) to discover what's worth testing, and output a CSV of candidate sanity-test scenarios. Security hygiene Check basic web security hygiene using a real browser — HTTPS enforcement and mixed content, cookie security flags (Secure/HttpOnly/SameSite), common security response headers (CSP, X-Content-Type-Opt Seo audit Check technical/structural SEO elements across a site using a real browser — title tags, meta descriptions, canonical tags, meta robots (catching accidental noindex), Open Graph tags, structured data Start testing Use whenever someone wants to test something but hasn't said (or doesn't know) which kind of check they need — "what can you test here", "what can you test on this site", "what can you do here", "what Test case writer Expand a terse scenario (from a scenario-mapper row, or a plain description) into a detailed, persistent E2E test-case document — preconditions, test data, numbered steps each with its own expected re Test plan Write or update a strategic test plan — scope, risk-based priorities, cadence for each of the other skills in this toolkit, release exit criteria, environment/data conventions, and a coverage ledger o User guide Generate an end-user-facing how-to guide for a site feature by walking through a verified flow with a real browser, capturing a screenshot with the relevant element visually highlighted at each step, Visual regression Compare two runs of visual-snapshot's screenshot capture — a baseline and a current run — using pixel-level diffing (pixelmatch, or @playwright/test's native toHaveScreenshot if a real Playwright Test Visual snapshot Navigate a site's main pages using a real browser (via Playwright MCP or @playwright/cli), capture full-page screenshots under consistent conditions, and compile them into a single reviewable gallery 