Test a website in a browser with AI
Ask an agent to follow a realistic user journey, capture the browser at the point of failure and tell you which step it could not complete.
Test a website in a browser with AI when a report such as “checkout is broken” leaves too much to guess. Give the agent one starting page and one visible sign of success: the order review opens, for example, or the form shows an error beside the empty field.
With playwright-cli the agent can inspect what is on the page before deciding which control to use. A screenshot and a few recorded actions can make a useful bug report even before anyone has opened the source code.
What it does
- Rehearse checkout up to the final confirmation.
- Check whether a required field has a visible error.
- Compare the same menu at two browser widths.
- Capture the state of a page after a click.
- Repeat a reported problem in a fresh browser session.
Install and use
- Provide an authorised test site, a supported browser and Microsoft’s Playwright CLI from npm. Installation needs Node.js 20 or later. Supply the complete skill folder and references. Use a separate test profile.
- Try: “Open the returns form on a narrow screen. Leave the order number empty, press Continue and tell me where the error appears.”
- Ask for the observed result alongside the expected one. Check screenshots for account details before sharing them.
FAQ
Source
Made by Microsoft. Original skill: playwright-cli, published on GitHub under the Apache-2.0 licence.
View on GitHub