skills.rest
mengblom
mengblomCOMMUNITY
Explore all skills by this author →
GitHub

agent-browser

Automate browser interactions via accessibility snapshot refs and Bash commands.

Updated Aug 27, 2026
Search Tags:#screenshots#browser automation#ai agents#web scraping#form filling#cli automation#accessibility snapshot
One-click install
npx skills add https://github.com/mengblom/dex-ghx --skill agent-browser-mengblom
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-browser
Source: https://github.com/mengblom/dex-ghx/tree/main/.claude/plugins/compound-engineering/skills/agent-browser
Command: npx skills add https://github.com/mengblom/dex-ghx --skill agent-browser-mengblom

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Browser automation becomes slow and error-prone when you try to manually navigate sites, fill forms, click controls, and capture evidence for an AI workflow.

Core Features & Use Cases

  • Ref-based page interaction: Uses accessibility snapshots and refs (such as @e1, @e2) to let an AI reliably click, fill, select, and wait on dynamic UI elements.
  • Form handling and data capture: Supports common actions like logging in, searching, extracting text/HTML/attributes, and taking screenshots or PDFs for documentation.
  • Scrape-style workflows with re-snapshotting: Re-snapshot after DOM changes to keep interactions aligned with the current page state.

Quick Start

Tell the agent to open https://example.com, take an interactive snapshot to find the relevant button ref, click it, and then capture a full-page screenshot.

Frequently Asked Questions about agent-browser

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

FAQPage Schema
How do I automate browser tasks like form filling and web scraping with AI agents?▼

AI-driven browser automation uses accessibility snapshot refs like @e1 to interact with web pages, supporting tasks like filling forms, clicking buttons, scraping content, and taking screenshots deterministically via CLI commands.

What are accessibility snapshot refs and how do they work for web automation?▼

Accessibility snapshot refs are identifiers assigned to page elements during a snapshot, allowing an AI agent to reliably target and interact with dynamic UI elements by executing deterministic Bash commands and re-snapshotting after DOM changes.

Do I need to install any CLI tools to use ref-based browser automation?▼

Yes, ref-based browser automation requires installing the agent-browser CLI tool from Vercel, which serves as the interface for taking accessibility snapshots and executing interaction commands on web pages.

Can I take screenshots and generate PDFs during automated web browsing sessions?▼

Yes, automated web browsing sessions support producing full-page screenshots and PDFs for documentation and evidence capture, alongside extracting text and HTML attributes from the current page state.

Why does my AI agent lose track of elements after clicking a button on a dynamic page?▼

Dynamic pages change the DOM after interactions like button clicks, requiring a re-snapshot to capture the updated page state and assign new refs so the AI agent can accurately target subsequent elements.

What's the best way to handle dynamic UI elements when scraping web pages with an AI agent?▼

The best approach is using ref-based selection from accessibility snapshots to interact with dynamic UI elements, waiting for specific elements or text to load, and re-snapshotting after navigation or DOM changes to maintain alignment.

Related Skills

Explore More Software Engineering Skills→
marcusrbrownmarcusrbrown
23

agent-browser

Automate browser interactions using Vercel's agent-browser CLI and ref-based selectors.

Community
Intermediate
Blake-JohnBlake-John

pymupdf

Convert PDF documents into Markdown, JSON, and text for LLM workflows.

Community
Intermediate
NourcineAbNourcineAb

browser

Automate web browser interactions, navigation, and data extraction for AI agents.

Community
Intermediate
Wike-CHIWike-CHI
2

agent-browser

Automate browser interactions and extract structured page content via CLI.

Community
Intermediate