Browser Control

To automate web navigation and form interaction, you select an open browser by its id and use typed references to click, type, and submit UI elements. The skill settles internal page state reads to ensure actions target the correct page reliably.

An agent inspects UI elements and verifies post-action changes by capturing fresh content and DOM state with typed reference markers. After an action like a click or submit, it reads the updated validation and state signals to confirm the expected UI change.

The best way to handle dynamic page content is using state-settling reads that capture fresh content and typed refs before acting. This ensures your click, type, or scroll operations target current elements, returning an action and state result to support controlled retries.

Yes, you can execute custom JavaScript using an evaluate and send escape hatch when selectors and typed refs cannot express a required browser action. This provides a last-resort method to interact with the page while still returning structured action status handling.

You manage multiple open browsers by listing all currently open browser instances and selecting your target by its numeric browser id. This directs your subsequent navigation, content reads, and UI actions to the correct page.

Browser automation fails to target correct UI elements after a page reload if the internal state is not settled. You must use content, detail, and dom helpers that internally settle navigation and capture fresh typed ref markers to ensure reliable follow-on actions.