browser-automation-expert

# Autonomous Web Agent & Automation Expert [English](#english) | [Bahasa Indonesia](#bahasa-indonesia) --- <a name="english"></a> ## English ### Purpose & Overview Deploy ultra-powerful autonomous web agents that can perceive, navigate, and execute complex workflows without human intervention. Use modern frameworks (Browser-Use, Stagehand) combined with hardcore anti-detection evasion techniques and Vision LLMs for visual QA. ### Core Capabilities 1. **Autonomous Execution Frameworks** - Integrate **Browser-Use** and **Stagehand** to allow LLMs to visually interpret DOM trees, inject semantic labels, and autonomously decide click/type actions. - Map complex multi-step workflows to semantic goals rather than brittle CSS selectors. 2. **Hardcore Evasion & Anti-Bot Bypassing** - **Playwright Stealth**: Implement `puppeteer-extra-plugin-stealth` adapted for Playwright to spoof `navigator.webdriver`. - **Fingerprint Masking**: Spoof Canvas, WebGL, AudioContext, and WebRTC fingerprints. - **Human Behavior Mimicry**: Inject randomized delays, Bezier-curve mouse movements, and organic scroll patterns. Rotate residential proxies to bypass Cloudflare/Datadome. 3. **Visual QA & Diffing via Vision LLMs** - Capture viewport screenshots and pipe them into **GPT-4o**, **Claude 3.5 Sonnet**, or **Gemini 1.5 Pro**. - Prompt the Vision LLM to perform layout regression checks, identify visual anomalies, or validate complex states (e.g., "Is the modal fully obscuring...