freeman-browser_skill

Overview

This skill provides a stealth, human-like browser wrapper that replaces raw Playwright, fetch, or curl for any web automation. It launches a fingerprinted Chromium session with realistic touch, mouse, typing, and timezone behaviors so sites rarely detect automation. Use it whenever you need robust access to protected or modern web apps without being blocked.

How this skill works

launchFreeman() is a drop-in replacement for chromium.launch() that returns a browser page plus human interaction helpers (humanType, humanClick, humanScroll, humanRead, sleep). It applies a default mobile Safari fingerprint, realistic input events, and optional desktop mode. The skill also exposes shadow DOM traversal, rich-text paste methods, CAPTCHA solving integration, request interception, and session persistence utilities.

When to use it

• Scraping or researching sites protected by Cloudflare, DataDome, PerimeterX, or other anti-bot systems
• Automating logins, form fills, button clicks, file uploads, and complex interactions on modern apps
• Accessing social platforms (Instagram, LinkedIn, X/Twitter, Reddit, TikTok) and e-commerce sites (Amazon) without being flagged
• Interacting with shadow DOM elements, web components, or contenteditable rich text editors
• Monitoring prices, tracking changes, or running repeatable browsing tasks with session persistence

Best practices

• Start with mobile default (iPhone 15 Pro) and switch to desktop ({ mobile: false }) for sites that reject mobile UAs
• Use humanType, humanClick, and humanScroll helpers instead of direct page.fill()/page.click() to improve reCAPTCHA and bot scoring
• Add deliberate waits (humanRead, sleep) and mouse entropy after navigation to satisfy Cloudflare-like checks
• Save and restore cookies or context to avoid repeated logins and reduce bot-detection signals
• When inputs return zero elements, check for Shadow DOM and use the provided deep query/fill utilities

Example use cases

• Login to Reddit using desktop mode, human typing delays, and Enter key submission to trigger correct form flow
• Scrape reviews or prices from Amazon while avoiding rate limits and bot blocks by using realistic browsing patterns
• Fill forms inside Shadow DOM or web components that normal selectors miss and extract submitted data
• Bypass Cloudflare checks by letting challenges resolve, adding human-like mouse/scroll entropy, and using appropriate waits
• Persist session cookies to automate repeat tasks without re-authentication between runs