agent-browser

Runs browser smoke tests, screenshots, and UI validation against forwarded localhost URLs.

Updated Jun 8, 2026
One-click install
npx skills add https://github.com/Solizardking/clawd-cloud --skill agent-browser-solizardking
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-browser
Source: https://github.com/Solizardking/clawd-cloud/tree/main/clawd-grok/.agents/skills/agent-browser
Command: npx skills add https://github.com/Solizardking/clawd-cloud --skill agent-browser-solizardking

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Verifying that a web app actually works after booting it in a sandbox requires a real browser, but the browser tooling lives on the host rather than inside the sandbox. This Skill bridges that gap by driving the host-side agent-browser CLI against forwarded localhost URLs. ## Core Features & Use Cases - Browser Smoke Tests: Open a forwarded localhost URL, wait for network idle, and snapshot the page to confirm the app renders. - UI Interaction & Validation: Click elements, fill inputs, and press keys using snapshot refs, then re-snapshot after DOM changes. - Evidence Capture: Record smoke tests as video and save screenshots to stable workspace paths under .grok/verify-artifacts/. - Use Case: After starting a dev server on port 3000 in the sandbox, open http://127.0.0.1:3000 in a named browser session, verify the page title and one critical button, and save a full-page screenshot as proof. ## Quick Start Use agent-browser to open the forwarded localhost URL of the running app, wait for network idle, take an interactive snapshot, and save a screenshot to .grok/verify-artifacts/.

Frequently Asked Questions about agent-browser

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

FAQPage Schema
How do I run a browser smoke test against a localhost app?▼

Start the app in the sandbox, forward its port, then run agent-browser with a named session to open the localhost URL, wait for network idle, and take an interactive snapshot. Keep checks narrow: page load plus one or two critical controls.

How do I capture screenshots with agent-browser?▼

Use the screenshot command with the --screenshot-dir flag pointing to a workspace-relative path like .grok/verify-artifacts/, not a positional path. Add --full for a full-page capture, and create the directory first with mkdir -p.

Can agent-browser run inside the sandbox environment?▼

No, agent-browser runs on the host, not inside Shuru. The workflow is to boot the app with sandboxed bash, expose it through a forwarded localhost port, and drive agent-browser against that URL from the host.

Why do my element refs stop working after a click?▼

Snapshot refs become stale after any click or navigation because the DOM changes. Re-run snapshot -i after each meaningful DOM change before using refs again.

What should I do if agent-browser is not installed?▼

Fall back to bash-only verification and state clearly that browser checks were skipped. The Skill requires the agent-browser CLI to be installed and available on PATH.