Dev Browser is an agent skill for the dev-browser command line tool. The assistant sends short JavaScript scripts to a warm background daemon that controls Chrome through Puppeteer, so a page opened in one run is still there in the next.
The skill covers installing the CLI, a quick start with named pages and accessibility snapshots, and the gotchas that trip up scripts: semicolons, refs that reset on navigation, and clicks that do not wait.
What You Get
- Install commands for the dev-browser CLI
- A quick start script with named pages and snapshots
- A list of gotchas for writing reliable scripts
Common Use Cases
- Navigating sites and filling forms
- Taking screenshots and scraping page data
- Testing web apps and logging into sites
What Makes It Unique
It keeps a warm daemon with pages that persist, so the agent does not restart a browser for every step, and it points the assistant to the CLI’s built-in help for the current API. The browser can be launched as an isolated profile or attached to one you already have open.
Under The Hood
The skill is one short SKILL.md with no scripts. The assistant runs the dev-browser command in a shell, which starts or attaches to a Chrome browser, loads any website you point it at, and can fill forms and click. The install step runs npm install -g and may download a binary.