playwright-clilisted

# Playwright CLI — disk-based frontend QA One bash call, one compact text report, screenshots on disk. Replaces the `browser_navigate` -> `browser_snapshot` -> `browser_take_screenshot` -> repeat MCP loop for the common case: a site whose DOM you already control (your own build, a scaffold, any local dev server) and you just need to confirm it renders clean. **Why this exists over MCP:** MCP's snapshot-per-step loop pays a full accessibility-tree or screenshot round-trip in tokens for every single action. A scripted Playwright run does the whole navigate-render-capture cycle for every route x viewport in one process and returns a handful of text lines. Screenshots land on disk — `Read` the PNG directly instead of paying for an MCP image round-trip. **When to still use Playwright MCP instead:** driving a page you don't control the markup of (a competitor site, a social-profile scrape, any external SaaS UI), or anything that needs live interaction — click a button, fill a form, drag, handle a dialog, and see what happens next. This skill only navigates and observes; it does not interact. ## Setup (one-time) ```bash cd "${CLAUDE_PLUGIN_ROOT}/skills/playwright-cli" && npm install ``` Pinned to `playwright@1.58.0`. If you already have the `playwright` CLI installed globally at the same version, this reuses its already-downloaded Chromium (`~/Library/Caches/ms-playwright/` on macOS) instead of pulling a second copy. ## Run it ```bash node "${CLAUDE_PLUGIN_ROOT}/skills/playw