browser-task

Browser automation context loss is prevented by delegating multi-step Chrome workflows to an isolated subagent, which executes visual QA and form testing independently before returning concise plain-text reports to the main thread.

A context-safe approach for multi-page navigation uses an isolated subagent to handle browser interactions, preserving the main conversation by relying on text-first inspection and controlled screenshot use rather than managing page state inline.

Delegate admin panel checks by handing off the browser workflow to a separate subagent, which performs text-first element reads and console log inspections, then reports the results back in plain text to preserve conversation history.

Browser-task does not require screenshots for every inspection; it applies a text-first workflow that prefers page text, element reads, searches, console logs, and network inspection before using high-cost visual screenshots.

You can use subagent delegation for contact form submission testing, which safely offloads the multi-step browser interactions and provides continuation support when tasks exceed limits during the extended Chrome workflow.

When a multi-step browser workflow exceeds subagent limits, the skill provides continuation support to resume the isolated browser task, ensuring the main conversation context remains preserved throughout the extended automation process.