lackeyjb / playwright-skill

Ask it to test a page, check responsive layouts, fill out a form, follow a login flow, or find broken links, and Claude writes custom Playwright code for that specific request, runs it through a bundled run.js executor that handles module resolution, and returns console output plus screenshots. The browser opens visibly by default (headless: false, 100ms slow motion) so you can watch the automation run rather than reading a pass/fail log after the fact; screenshots land in /tmp/ unless you ask otherwise. A separate API_REFERENCE.md covering selectors, network interception, auth, visual regression, mobile emulation, and performance testing loads only when the task needs it, rather than being read on every invocation.

It's a general-purpose browser-testing skill for anyone using Claude Code who wants ad hoc verification of a website — "test if google.com loads," "check the contact form" — rather than a fixed regression suite you maintain yourself. Because Claude writes the automation on the spot each time, there's no library of named test cases to browse or extend; each run is a one-off script generated for what you asked.