agent-browser

Automate browser interactions by orchestrating a CLI-based browser driver to open pages, snapshot elements with refs, and perform actions like click and fill. You capture a page snapshot to locate elements, then instruct the agent to interact using those refs to complete tasks such as form automation.

Snapshot-based ref identification is a mechanism where the browser captures a page state and assigns refs to elements. The agent uses these refs to accurately locate and interact with UI components for web testing or data extraction without relying on brittle CSS selectors.

Yes, you can maintain state across multiple sessions for automated UI testing. The skill supports multi-session browsing across sites with optional state persistence, allowing you to continue interactions like dashboard login flows across different automation runs.

Yes, a CLI-based browser driver is required to extract data from web catalogs. The skill depends on this driver to open URLs, capture page snapshots, and use refs to navigate and extract product data automatically.

The best way to automate login flows is to open the dashboard URL, snapshot the page to capture element refs, and use those refs to fill credentials and click submit. You can then take a screenshot or snapshot to validate the resulting UI state.

Browser automation fails to locate elements if you attempt interactions before capturing a page snapshot. You must snapshot the page after it loads to generate fresh refs, as the agent relies exclusively on these refs to identify and interact with UI elements accurately.