A high-performance headless browser automation CLI that provides AI agents with deterministic element selection through structured accessibility tree snapshots.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install browser-automation-clawdbot
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install browser-automation-clawdbot using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Browser Skill?
The Agent Browser Skill is a developer-centric automation tool optimized for Openclaw Skills. It solves the common problem of non-deterministic web interaction by using accessibility tree snapshots instead of visual pixels or fragile DOM selectors. By providing agents with stable references (refs) for every interactive element, it ensures that automation workflows remain robust even on complex Single Page Applications (SPAs). This skill is essential for developers building agents that need to navigate, interact with, and extract data from the modern web with high precision and low latency.
Agent Browser Skill Use Cases
- Automating complex, multi-step web workflows that require high reliability
- Interacting with dynamic SPAs where traditional CSS selectors are unstable
- Managing multiple isolated browser sessions for parallel agent tasks
- Bypassing repetitive login flows by persisting authentication states across sessions
- Performing structured data extraction from web pages via accessibility metadata
How Agent Browser Skill Works
- The agent opens a browser instance and navigates to a target URL using a specific session context.
- It generates a snapshot with the -i and --json flags to retrieve a structured view of all interactive elements.
- The agent parses the accessibility tree to identify target elements and their associated refs (e.g., @e1, @e2).
- Interaction commands like click, fill, or scroll are sent directly to the identified refs for deterministic execution.
- The agent monitors the page state or network idle status before taking subsequent snapshots to continue the workflow.
Agent Browser Skill Setup
To integrate this tool into your Openclaw Skills workflow, install the CLI and the necessary browser binaries using the following commands:
npm install -g agent-browser
agent-browser install
# On Linux, include system dependencies
agent-browser install --with-deps
Agent Browser Skill Data Schema & Taxonomy
The skill organizes its output into structured JSON objects, making it easy for AI agents to process page state. The primary snapshot schema is described below:
| Key | Description |
|---|---|
| success | Boolean indicating the result of the CLI operation |
| data.snapshot | A human-readable text representation of the page accessibility tree |
| data.refs | A map of element IDs to metadata including roles (button, link) and names |
| data.url | The current URL of the active browser tab |
Agent Browser Skill Advanced Features
- Named sessions for isolating multiple browser contexts and preventing state leakage
- State persistence allowing you to save and load cookies or local storage for session recovery
- Network routing capabilities to block unnecessary resources like ads or mock API responses
- Advanced frame and tab management for navigating complex web applications with iframes
- Deterministic waiting conditions based on text, URLs, or custom JavaScript functions
SKILL.md
Loading
Related Openclaw Skills
Alibaba Cloud NIS Network Reachability Analysis
An expert diagnostic tool for performing deep network path analysis and troubleshooting across Alibaba Cloud resources using Network Intelligence Service (NIS).
Weiqi Joseki Database
An automated, local-first database tool for extracting and analyzing high-frequency Go joseki from KataGo AI game archives.
Find Skills
A discovery and management tool to find and install specialized agent capabilities from the Openclaw Skills ecosystem.
Mapbox iOS Integration Patterns
A comprehensive implementation guide for integrating Mapbox Maps SDK v11 into iOS applications using Swift and SwiftUI.
Skill Vetter
A security-first vetting protocol designed to analyze and verify the safety of AI agent skills before installation.
Voice Call
A comprehensive plugin enabling AI agents to initiate, manage, and interact via voice calls using major telephony providers.
Featured*
