How does the agent loop function?

The agent captures a screenshot, sends it to the model to determine the next action, executes that action via Playwright, and repeats the process until the task is complete.

Does it support safety features for browser actions?

Yes, the skill includes a safety_decision logic that prompts the user for confirmation before executing actions flagged as risky or sensitive.

Which browsers can I use with this skill?

By default, it uses Playwright's bundled Chromium, but it also supports Google Chrome, Microsoft Edge, and custom Chromium-based executables like Brave.

What is the Gemini Computer Use skill?

It is a specialized capability that allows Claude to create and run agents that control a web browser using the Gemini 2.5 model and Playwright library.

Gemini Browser Automation

Name: Gemini Browser Automation
Author: sarfraznawaz2005

bysarfraznawaz2005

•

Browser Automation

Builds and executes autonomous web agents using Gemini 2.5 and Playwright to automate complex browser-based workflows.

The Gemini Computer Use skill enables developers to create sophisticated browser-control agents that leverage Gemini 2.5's vision and reasoning capabilities. It provides a structured agent loop that captures screenshots, analyzes them through the model, and executes precise navigation or interaction actions via Playwright. This skill is ideal for automating multi-step web tasks, performing visual research, or managing browser-based tools while maintaining security through human-in-the-loop safety confirmations and sandboxed execution.

Key Features

01Customizable viewport settings and turn-limit controls

02Support for sandboxed profiles and excluded action lists

03Multi-browser support including Chromium, Chrome, Edge, and Brave

04Integrated human-in-the-loop safety confirmation for risky UI actions

05Autonomous agent loop for screenshot analysis and action execution

061 GitHub stars

Use Cases

01Automating repetitive web research and data gathering tasks

02Creating browser-based agents for complex form completion and workflows

03Conducting automated UI/UX walkthroughs and visual testing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sarfraznawaz2005/agent-skills-collection gemini-computer-use

For use in Claude.ai and ChatGPT

Download Skill