browse

Controls a persistent headless Chromium browser for QA testing, screenshots, and page state verification.

1|Updated Jul 23, 2026
One-click install
npx skills add https://github.com/sanjanb/my-agent-harness --skill browse-sanjanb
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: browse
Source: https://github.com/sanjanb/my-agent-harness/tree/main/skills/gstack-browse
Command: npx skills add https://github.com/sanjanb/my-agent-harness --skill browse-sanjanb

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bun.

What problem does it solve? Manually verifying that a web feature works—clicking through flows, checking console errors, capturing screenshots for bug reports—is slow and hard to reproduce. This Skill gives an AI agent a persistent headless Chromium session so it can navigate pages, interact with elements, and collect visual and console evidence automatically. ## Core Features & Use Cases - Interactive QA flows: Navigate URLs, fill forms, click elements, upload files, handle dialogs, and assert element states (visible, enabled, checked, focused) at roughly 100ms per command after startup. - Visual evidence and diffs: Take annotated or plain screenshots, capture responsive layouts across mobile/tablet/desktop viewports, and diff page snapshots before and after an action to verify exactly what changed. - Diagnostics: Inspect console errors, network failures, and page text, compare staging versus production environments, and render local HTML files without running a server. - Use Case: After shipping a login flow, ask the agent to open the site, fill in credentials, submit the form, diff the resulting page, and attach an annotated screenshot plus console log as proof the flow works. ## Quick Start Ask the agent to open your app in the browser, test the login flow, and show you a screenshot of the result.

Frequently Asked Questions about browse

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I test a website with a headless browser?

Start with the goto command to load a URL, then use text, console, and network commands to verify content loads without errors. Use snapshot -i to list interactive elements, then fill and click to walk through user flows.

How do I take annotated screenshots for a bug report?

Run snapshot -i -a -o with an output path to capture a labeled screenshot of interactive elements, or use screenshot for a plain capture. Combine with the console command to attach error logs as evidence.

Can I test responsive layouts across mobile and desktop?

Yes. The responsive command captures mobile, tablet, and desktop screenshots in one step, and viewport lets you set an exact size like 375x812 before taking a screenshot.

Does the headless browser keep login sessions between commands?

Yes. State persists between calls, including cookies, tabs, and login sessions. The first call auto-starts the browser in about 3 seconds, then subsequent commands run in roughly 100ms.

Why does the browse command need a setup step first?

The browser binary must be built once via the setup script, which requires bun. If bun is missing, the skill installs it with a checksum-verified installer before compiling.

Can I render local HTML files without running a server?

Yes. Use goto with a file:// URL for files on disk, or load-html to inject HTML content directly. Both are scoped to files under the working directory or the temp directory.