Browser automation
Lets your agent open a real browser from the terminal to click through a flow, fill forms, take screenshots and read what’s on the page, so it can show you a change working.
When to use it
- You want proof that a sign-up, checkout or settings flow still works after a change.
- A layout problem only shows up at one screen size or after a few clicks.
- You need screenshots or page text captured the same way every time.
If you want a permanent test suite in the repository, ask for Playwright test files instead, because this skill drives the browser from the command line.
A practical example
You ask
“Open the sign-up page at phone width, fill the form with test values and send me a screenshot of the confirmation screen.”
Your agent does
It opens a browser through the bundled wrapper, steps through the form, saves the screenshot and reports anything that overlapped or clipped.
What gets installed
The whole folder is copied, so scripts, references and any license notices stay together.
- agents/openai.yaml353 bytes
- assets/playwright-small.svg828 bytes
- assets/playwright.png1,730 bytes
- LICENSE.txt11,351 bytes
- NOTICE.txt403 bytes
- references/cli.md1,863 bytes
- references/workflows.md1,864 bytes
- scripts/playwright_cli.sh526 bytes
- SKILL.md4,973 bytes
Before you start: Needs Node.js with npx, which the bundled wrapper uses to run playwright-cli.
Credit and license
Adapted from Microsoft’s playwright-cli skill, including its NOTICE.txt attribution. It’s shared under the Apache License 2.0, and the license file is installed with it however you choose to install.