host-browser

Drive the user's real logged-in Chrome on macOS via the agent-browser CDP bridge.

Updated May 3, 2026
One-click install
npx skills add https://github.com/spikelab/multiplai-cc-mktplace --skill host-browser-spikelab
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: host-browser
Source: https://github.com/spikelab/multiplai-cc-mktplace/tree/main/plugins/multiplai-media/skills/host-browser
Command: npx skills add https://github.com/spikelab/multiplai-cc-mktplace --skill host-browser-spikelab

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Automated browsers get blocked: ephemeral test browsers leak navigator.webdriver=true, lack cookies and logins, and trigger bot walls, CAPTCHAs, and 403s on JS-heavy or protected pages. This Skill attaches automation to the user's real, persistent, logged-in Chrome on a macOS host so sessions look and behave genuinely human. ## Core Features & Use Cases - Real-Chrome attachment: hb-connect.sh binds the agent-browser daemon to the host Chrome over CDP (port 9222) and asserts navigator.webdriver === false before any action. - Human-paced interaction verbs: the hb wrapper adds humantype, humanclick (curved, jittered mouse paths), think, fillform, waitfor, and consent-overlay dismissal via hb goto/hb dismiss. - Risk-scored wall handling: hb fpcheck, hb dd, hb warmup, and hb solve-wait detect DataDome/PerimeterX-style scoring and hand any rendered challenge to the human rather than auto-solving it. - Structured data extraction: hb data reads embedded __NEXT_DATA__/__NUXT__/JSON-LD blobs instead of scraping virtualized DOM lists. - Use Case: Sign up for a service that blocks headless browsers — open a mail.tm temp inbox with hb mail new, fill the signup form with human-paced typing, retrieve the verification code with hb mail code, and complete the flow from the genuine browser session. ## Quick Start Run the hb-connect.sh script to attach to my real Chrome, then use hb goto to open the bot-blocked page and read its content.

Frequently Asked Questions about host-browser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate Chrome without getting detected as a bot?

Attach to a real, normally-launched Chrome via CDP instead of launching a test browser. This keeps navigator.webdriver false, preserves genuine cookies and fingerprint, and combined with human-paced typing and clicking passes most behavioral detection.

How to bypass DataDome or PerimeterX bot protection?

These systems score risk rather than gate every request. Stay under the threshold with a residential IP, warm profile navigation, and low request velocity; if a challenge renders, the human solves it manually and the earned datadome cookie is reused.

Why does my scraper only extract a few items from a results page?

Modern SPAs virtualize long lists, mounting only 2-10 visible rows in the DOM. Read the embedded JSON state instead — __NEXT_DATA__, __NUXT__, or JSON-LD blobs contain the full dataset via the hb data command.

Can I use this browser automation on Linux or without a container?

The skill targets a macOS host reached through a container-to-host SSH bridge running agent-browser. Running directly on the Mac works too, since CDP on port 9222 just needs to be reachable locally.

Why do my automated clicks and typing do nothing on a page?

A cookie consent or modal overlay is likely swallowing the clicks. Run hb dismiss or arrive via hb goto to accept cookies and close pop-ups, then re-snapshot to get fresh interactive element refs.

What are the ethical limits of anti-detection browser automation?

Authorized, human-rate use only: no fingerprint spoofing, no CAPTCHA auto-solving, no proxy rotation, no bulk account creation. Explicit policy walls like disposable-email blocks or no-automation ToS are stop signals, not puzzles to defeat.