skills.rest
MinhHoangDono
MinhHoangDonoCOMMUNITY
Explore all skills by this author →
GitHub

ck:agent-browser

Automate browser interactions using snapshot-based element references for AI agents.

Updated Mar 7, 2026
Search Tags:#browser automation#ai agent#ui testing#recording#session persistence#snapshot refs#cloud testing
One-click install
npx skills add https://github.com/MinhHoangDono/antigravity-kit --skill ck-agent-browser-minhhoangdono
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ck:agent-browser
Source: https://github.com/MinhHoangDono/antigravity-kit/tree/main/.agent/skills/agent-browser
Command: npx skills add https://github.com/MinhHoangDono/antigravity-kit --skill ck-agent-browser-minhhoangdono

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It solves the challenge of running reliable, context-efficient browser automation from an AI agent without manually scripting brittle UI selectors or overwhelming the model with full-page context.

Core Features & Use Cases

  • Context-efficient “snapshot + refs”: capture the accessibility tree and interactive elements, then act using stable reference IDs (e.g., @e1, @e2) across a long workflow.
  • Interactive automation commands: navigate, fill, click, type, hover, select, upload files, and re-snapshot after each state change to keep actions self-correcting.
  • Debugging and test support: record videos, export screenshots/PDFs, manage cookies/state persistence for auth, and run parallel sessions (optionally on Browserbase for cloud testing).
  • Use case: Use it to complete a multi-step sign-in and dashboard flow end-to-end by taking an initial interactive snapshot, performing actions by refs, waiting for URL/state changes, and re-snapshotting until the workflow verifies success.

Quick Start

Tell your AI to run: open https://example.com, take an interactive snapshot, click the referenced sign-in button, then re-snapshot and wait for the dashboard URL.

Frequently Asked Questions about ck:agent-browser

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

FAQPage Schema
How do I automate browser interactions with an AI agent without brittle selectors?▼

Browser automation uses context-efficient snapshot-based element references for AI agents. You capture the accessibility tree, then execute actions like click or fill using stable reference IDs, keeping workflows self-correcting without manual selectors.

What is the snapshot and ref workflow for end-to-end UI testing?▼

The snapshot and ref workflow involves opening a URL, taking an interactive snapshot to capture interactive elements, performing actions by reference IDs, then re-snapshotting and waiting for state changes. This loop verifies multi-step dashboard flows end-to-end.

Can I run parallel browser sessions and record media for CI testing?▼

Yes, you can run parallel browser sessions, record videos, export screenshots/PDFs, and optionally execute cloud testing on Browserbase. These features support debugging and CI testing for long autonomous multi-tab workflows.

Does browser automation support multi-tab workflows and session persistence?▼

Browser automation supports multi-tab workflows, long autonomous sessions, and robust browser state management. It handles cookies and state persistence to maintain authentication across extended interactive automation tasks.

Why does my AI agent lose context during long browser automation workflows?▼

AI agents lose context when overwhelmed with full-page context or brittle selectors. Using context-efficient snapshot-based element references provides stable interactive IDs, preventing context loss across long autonomous browser workflows.

Do I need Browserbase to run cloud testing for browser automation?▼

No, Browserbase is an optional dependency for cloud execution in CI testing. The browser automation can run locally using the snapshot-based reference workflow for interactive element actions and UI verification loops.

Related Skills

Explore More Software Engineering Skills→
ProyectoG007ProyectoG007

agent-browser

Automate headless browser workflows using accessibility tree snapshots for element selection.

Community
Advanced
viably-labsviably-labs

playwright-screenshot-inspector

Analyze Playwright screenshots to validate UI rendering accuracy semantically.

Official
Advanced
dthuy62dthuy62

ck:agent-browser

Automate browser navigation, inspection, and interaction workflows with snapshot-driven refs.

Community
Advanced
idan-rubinidan-rubin
41

browserclaw

Automate browser interactions via text snapshots and numbered refs.

Community
Intermediate