A high-performance headless browser CLI designed for AI agents using accessibility tree snapshots for deterministic element interaction.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install wayne-agent-browser
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 wayne-agent-browser 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?
Agent Browser is a specialized automation tool within the Openclaw Skills ecosystem designed specifically for the unique needs of AI agents. Unlike traditional scrapers, it utilizes accessibility tree snapshots with unique reference IDs (refs), allowing agents to select and interact with elements deterministically without relying on fragile CSS selectors or XPaths. This approach ensures high reliability and speed when navigating complex single-page applications (SPAs).
By focusing on the accessibility layer, Agent Browser provides a machine-readable representation of the web that mirrors how humans perceive interactive elements. As a core part of the Openclaw Skills collection, it empowers developers to build agents that can execute multi-step workflows, manage isolated browser sessions, and maintain state persistence across complex tasks.
Agent Browser Use Cases
- Automating multi-step web workflows that require high reliability.
- Navigating complex React or Vue-based single-page applications where element IDs change frequently.
- Running parallel browser tasks with isolated session contexts.
- Bypassing repetitive login flows by saving and loading authentication states.
- Deterministic data extraction from dynamic web content using AI-driven element identification.
How Agent Browser Works
- Navigation: The agent opens a target URL using the open command to initiate a browser instance.
- Snapshoting: The tool generates an accessibility tree snapshot (using the -i and --json flags) which maps interactive elements to specific refs like @e1 or @e2.
- Analysis: The AI agent parses the JSON snapshot to identify the correct reference ID for the intended action.
- Interaction: The agent executes commands such as click, fill, or hover directly against the identified refs.
- State Management: After page changes, the agent re-snapshots the page or waits for network idle to ensure the DOM is stable before the next interaction.
Agent Browser Setup
To get started with this component of Openclaw Skills, install the CLI globally via npm and download the necessary browser binaries:
npm install -g agent-browser
agent-browser install
# On Linux systems, you may need to install system dependencies
agent-browser install --with-deps
Agent Browser Data Schema & Taxonomy
Agent Browser organizes interaction data through a structured JSON output format during snapshots. This schema allows Openclaw Skills to maintain a clear map of the page's interactive landscape.
| Key | Type | Description |
|---|---|---|
success |
Boolean | Indicates if the snapshot or command was executed successfully. |
data.snapshot |
String | A text representation of the accessibility tree. |
data.refs |
Object | A dictionary mapping refs (e.g., @e1) to element metadata like role and name. |
data.refs[id].role |
String | The ARIA role of the element (e.g., button, heading). |
data.refs[id].name |
String | The accessible name or label of the element. |
Agent Browser Advanced Features
- Session Isolation: Manage multiple independent browser contexts simultaneously using the --session flag.
- State Persistence: Save and load cookies and local storage to auth.json to skip complex login sequences.
- Network Routing: Intercept, block, or mock network requests and API calls for testing and performance.
- Scope Control: Limit snapshots to specific DOM selectors to reduce token usage and improve AI processing speed.
- Interactive Mode: Use the --headed flag during development to visually debug agent interactions in real-time.
SKILL.md
Loading
Related Openclaw Skills
Minijuegos LeChuck SDK Integration
A comprehensive integration guide for the Minijuegos.com LeChuck JS SDK to handle player authentication, achievements, and persistent game statistics.
Content Repurposer
The Content Repurposer transforms long-form articles, blogs, and transcripts into punchy, platform-optimized social media posts for LinkedIn, Twitter, and Instagram.
Techwrite: Deep Technical Writing Workflow
Techwrite is a professional-grade documentation engine that transforms complex technical concepts into actionable, high-precision prose for developers and stakeholders.
Serverless Deep Workflow
A comprehensive six-stage framework for designing, debugging, and scaling resilient serverless workloads across AWS Lambda, Azure Functions, and Google Cloud Functions.
Browser Automation
A comprehensive automation suite for programmatic web interaction, data extraction, and cross-browser testing.
Industry Stock Tracker
An automated financial reporting tool that generates structured industry and stock analysis reports from natural language queries.
Featured*
