Show HN: Playwright Skill for Claude Code – Less context than playwright-MCP (github.com)

A new Claude Skill — packaged as a Claude Code plugin — gives Claude the ability to write and execute Playwright browser automations on demand. Ask Claude to test a page, fill forms, click through flows, capture screenshots or run responsive and performance checks, and it will generate custom Playwright code, run it locally via a universal executor (run.js) and return screenshots and console output. The skill is model-invoked (Claude decides when to use it), supports visible browsing by default (headless: false), and is distributed through the Claude plugin marketplace or manual GitHub installation. Technically, the skill emphasizes reliability and developer ergonomics: run.js handles module resolution to avoid “module not found” errors, smart temp-file cleanup prevents races, and optional helpers speed common tasks. Defaults include slowMo 100ms, 30s timeouts, and screenshots saved to /tmp. It requires Node.js >= 14, Playwright ^1.48.0, and Chromium (npm run setup installs dependencies and browsers). Detailed docs are progressively loaded—SKILL.md for quick usage and a 630-line API_REFERENCE.md for advanced features (selectors, network interception, auth, visual regression, mobile emulation, debugging). Significance: this lowers friction for turning natural-language test requests into executable end-to-end automations, enabling faster QA, prototyping, and programmatic browsing workflows — while also surfacing operational considerations around running generated code and local browser automation.