skills.rest
Ricardo-Vae
Ricardo-VaeCOMMUNITY
Explore all skills by this author →
GitHub

web-access

Coordinate search, fetch, and CDP browser automation for authenticated web research.

Updated May 13, 2026
Search Tags:#web research#cdp#web automation#file upload#media extraction#authenticated browsing#site experience
One-click install
npx skills add https://github.com/Ricardo-Vae/codex-research-skills --skill web-access-ricardo-vae
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: web-access
Source: https://github.com/Ricardo-Vae/codex-research-skills/tree/main/skills/web-access
Command: npx skills add https://github.com/Ricardo-Vae/codex-research-skills --skill web-access-ricardo-vae

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires node, sqlite3, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill enables AI agents to perform real web browsing and automation tasks with dependable routing across search, fetch, and full browser (CDP) interaction, including flows that require existing login state.

Core Features & Use Cases

  • Intelligent tool selection for web tasks: Automatically chooses between WebSearch/WebFetch/curl/Jina and full browser CDP based on what the job actually needs.
  • CDP-backed browser automation (login-aware): Uses a CDP Proxy connected to the user’s existing Chrome so the agent can operate dynamic pages, navigate, click, upload files, and extract content as if a real user is browsing.
  • Site experience accumulation: Saves verified per-domain interaction facts (patterns, traps, and platform characteristics) to improve future success across sessions.
  • Media extraction from live DOM/video: Extracts image/video URLs from the DOM and supports video frame sampling via screenshot after controlling the page.

Quick Start

Use the web-access skill to complete a logged-in, interactive web task by telling your agent: “Use web-access for this task and follow its workflow completely.”

Frequently Asked Questions about web-access

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

FAQPage Schema
How do I automate web browsing tasks that require an existing login state?▼

To automate authenticated web browsing, this Skill uses a CDP Proxy connected to your existing Chrome instance, allowing agents to operate dynamic pages, click, and upload files as a logged-in user. It automatically selects the right tool between search, fetch, and full browser interaction.

Can AI agents extract media like images and video frames from a live DOM?▼

Yes, agents can extract media from a live DOM by retrieving image and video URLs directly. For video content, the Skill supports frame sampling via screenshots after the agent controls the page through Chrome remote debugging.

What is the best way to handle dynamic web research across multiple sessions?▼

The best way to handle dynamic web research across sessions is using a Skill that accumulates site experience. This Skill saves verified per-domain interaction facts, patterns, and platform characteristics to improve future automation success.

Do I need Node.js and Chrome remote debugging to use CDP browser automation?▼

Yes, you need Node.js 22+ with Chrome remote debugging enabled to use CDP browser automation. The Skill relies on this environment plus its CDP Proxy API orchestration to ensure proper tool selection and page state verification.

How does intelligent tool selection work for fetching web content?▼

Intelligent tool selection for fetching web content works by automatically routing tasks between WebSearch, WebFetch, curl, Jina, and full browser CDP based on the job's requirements. This ensures efficient access to reliable sources and rendered pages.

Related Skills

Explore More Education & Research Skills→
ChromeDevToolsChromeDevTools
48.8k

chrome-devtools

Control Chrome DevTools via MCP for debugging and browser automation.

Official
Intermediate
Yapie0Yapie0
55

video-frames

Extract frames and thumbnails from video files using ffmpeg.

Community
Intermediate
viktorbezdekviktorbezdek
10

memory-systems

Implements cross-session AI memory using vector stores, graphs, and knowledge bases.

Community
Advanced
H2OSLabsH2OSLabs
1

openapi-to-components

Convert OpenAPI specifications into typed Next.js API clients and fetchers.

Official
Advanced