Route all networked work - search, scraping, logged-in actions, dynamic pages, and social platforms like Xiaohongshu or Weibo - through a real browser via a Chrome DevTools Protocol proxy.
---
name: web-access
description: Handle all online operations through a real browser - search, web scraping, logged-in actions, dynamic pages, and social platforms (Xiaohongshu, Weibo, Twitter). Use whenever a task needs a genuine browser environment.
---
A goal-driven web workflow that drives a real browser over the Chrome DevTools Protocol (CDP).
## Workflow
1. Pre-flight: run `scripts/check-deps.mjs` (needs Node.js 22+ for native WebSocket); follow its exit codes - continue, prompt for a browser preference into `config.env`, or self-heal per the printed steps. Switch browsers with `--browser chrome|edge` after killing the long-lived `cdp-proxy.mjs`.
2. Show the user the built-in automation-risk notice, then start the CDP proxy.
3. Browse like a human: define success criteria, pick the most direct starting point, and choose CDP directly for pages needing interaction, login state, or platforms known to block static fetches.
4. Treat each step's result as evidence - adapt, go deeper when content is thin, and keep deciding against the goal rather than a fixed script.
Full skill & source: https://github.com/eze-is/web-access/tree/7af34af6a25940d917905f0e5f2a7ef056952971