agent-browser-stealth for Openclaw

A stealth-first web automation skill for Openclaw Skills that enables AI agents to navigate bot-protected websites with human-like behavior.

leeguooooo
v1.0.2
Mar 2, 2026
0
824
7

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install agent-browser-stealth

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install agent-browser-stealth 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-stealth?

agent-browser-stealth is a robust automation skill designed to provide AI agents with a high-stability interface for interacting with the web. In the context of Openclaw Skills, this tool prioritizes anti-fingerprint evasion and stealth-safe interaction patterns, making it ideal for tasks involving sensitive platforms like Shopee, TikTok, or other e-commerce sites. It bridges the gap between simple script execution and human-like browsing by employing deterministic command flows and stable session management.

By utilizing this skill, developers can ensure their agents maintain persistence through login flows and navigate region-sensitive targets without triggering security triggers. The skill leverages a specific CLI package to handle complex tasks like captcha-prone flows and contenteditable editors, providing a reliable foundation for any browser-based agent activity.

agent-browser-stealth Use Cases

  • Automating purchases or data extraction on e-commerce platforms with strict anti-bot measures.
  • Managing persistent social media or account logins using session reuse.
  • Performing web research on region-locked sites by aligning locale and timezone settings.
  • Navigating complex web applications that require humanized typing pacing and interactive element snapshots.

How agent-browser-stealth Works

  1. The agent initializes a browser session using the agent-browser CLI, optionally specifying a session name for state persistence.
  2. A target URL is opened, and the tool waits for a specific load state, such as network idle, to ensure page stability.
  3. The agent triggers an interactive snapshot, which generates specific element references (e.g., @e1, @e2) for precise interaction.
  4. Commands like click, fill, or type are executed using these references, often with built-in delays to mimic human behavior.
  5. The agent performs a re-snapshot after each action to verify state changes and prepare for the next step in the workflow.

agent-browser-stealth Setup

To integrate this capability into your Openclaw Skills environment, install the stealth package and initialize the browser engine:

pnpm add -g agent-browser-stealth
agent-browser install
agent-browser --version

agent-browser-stealth Data Schema & Taxonomy

The skill manages data through several key structures to ensure reliability and traceability:

Component Description
Session State Persistent cookies and local storage saved under unique session names.
Element Refs Dynamic identifiers (e.g., @eN) mapped to DOM elements during a snapshot.
Interaction Logs Detailed records of executed commands, including URLs and element targets.
Visual Snapshots PNG screenshots generated for verification or debugging of the page state.

agent-browser-stealth Advanced Features

  • Session Persistence: Reuse named sessions to avoid repeated cold-start fingerprints and maintain login states.
  • Humanized Pacing: Customizable delays for typing and navigation to minimize bot signals.
  • Headed Mode Support: Ability to run in headed mode for bypassing the most sensitive bot protection layers.
  • Deterministic Element Targeting: Uses interactive snapshots to create stable references, reducing errors from brittle CSS selectors.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*