Tool name: Anthropic’s webapp-testing skill in the public anthropics/skills repository. The upstream source is unusually concrete about agent behavior. It tells the agent to write native Python Playwright scripts, use scripts/with_server.py to manage one or more local servers, and follow a reconnaissance-then-action workflow for dynamic apps. That means the skill is not just “use Playwright.” It is a repeatable operator playbook for verifying local web applications.

What the agent does: start the frontend and, if needed, backend servers; wait for ports to come up; open the rendered app in headless Chromium; wait for networkidle so JavaScript has actually settled; inspect the live DOM; capture screenshots; identify reliable selectors from rendered state; click, type, and assert expected behavior; and return logs or screenshots when something breaks. The helper supports single-server and multi-server setups, which matters for modern apps that split frontend and API services.

When to use it: invoke this skill when the job is to verify a local web app, reproduce a UI bug, smoke-test a new feature before merge, or debug why a dynamic page behaves differently after JavaScript runs. This is the right fit when the user should not have to manage server lifecycle, port readiness, and Playwright orchestration manually.

Scope boundary: this is not a generic Playwright product card, not a browser-grid listing, and not a catch-all web scraping framework entry. Its boundary is local webapp verification and debugging through the specific workflow defined in the upstream skill. Integration points: Python, Playwright, local npm or Python dev servers, screenshots, browser logs, and optional CI usage for repeatable smoke checks.