Web Access Skill - Setup Guide¶

Source: eze-is/web-access GitHub: eze-is/web-access Skills: 1 skill · 15.7K installs Category: Web Access & Scraping First Seen: catalogued August 17, 2026 sweep (web-access on skills.sh since March 19, 2026) Quality Tier: 🟡 Trusted - Socket Warn and Snyk Fail named (see Limitations)

web-access is a unified web access skill covering three methods in one workflow: WebSearch for discovery, WebFetch/curl for static content, and CDP browser automation for login-protected or dynamically-rendered platforms (Xiaohongshu, WeChat, Twitter, and similar). It drives your existing Chrome or Edge through a CDP proxy - preserving login state, running in background tabs, with no separate browser instance.

Installation¶

npx skills add eze-is/web-access

Individual skills install with the explicit repo form:

npx skills add https://github.com/eze-is/web-access --skill web-access

Prerequisites¶

Requirement	Details
Node.js 22+	Required - uses native WebSocket
Chrome or Edge	Your existing browser, driven via CDP proxy
Dependency check	Skill runs `check-deps.mjs` before any network operation

What It Provides¶

Three access methods in one skill: search discovery, static fetch, and CDP browser automation for login-gated content
10+ core commands: /eval (DOM queries and JS execution), /click (UI interaction), /screenshot (visual capture), /scroll (lazy-load triggering), /navigate (page transitions), plus file upload support
Human-like browsing doctrine: clarify success criteria first, choose the most direct path, validate results against goals, adjust on evidence rather than repeating failed approaches
Parallel sub-agent execution for independent research targets, with a shared browser instance and automatic tab cleanup

Quick Start¶

Install: npx skills add eze-is/web-access
The dependency check confirms CDP availability before any operation
Ask: "search for X and open the results in the existing browser session"

CorpusIQ Use Cases¶

Use Case	How
Login-gated research	Preserves real browser sessions - the same property our Mac Mini Playwright context provides, available to any agent
Resilient fetch path	A third fallback path when our primary search backends degrade
Social-platform checks	CDP interaction with platforms that block headless automation

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations / Verification¶

Security audits on web-access: Gen Agent Trust Hub Pass, Socket Warn, Snyk Fail - named per catalog policy; review before use in security-sensitive contexts
Publisher-page install counts verified (15.7K single skill); GitHub 8.7K stars on the repo
SKILL.md front matter is bilingual (Chinese primary) - the dependency check and config prompts appear in Chinese first
Requires a real Chrome/Edge session on the host - not a headless replacement

npx skills add eze-is/web-access   # verify install works

Powered by CorpusIQ