9.2K ·GitHub
What is this?
An open-source Agent Skill that gives AI coding agents full web access: it adds smart selection among web tools (WebSearch, WebFetch, curl, Jina, CDP) plus real browser automation through a proxy connected to your own browser.
An open-source Agent Skill that gives AI coding agents full web access: it adds smart selection among web tools (WebSearch, WebFetch, curl, Jina, CDP) plus real browser automation through a proxy connected to your own browser.Is this a standalone app or something my coding agent loads?It's a skill package — a SKILL.md plus Node.js scripts — that your agent loads and follows; it works with Claude Code, Cursor, Gemini CLI, Codex CLI and anything else supporting SKILL.md. You never run it on its own.
It's a skill package — a SKILL.md plus Node.js scripts — that your agent loads and follows; it works with Claude Code, Cursor, Gemini CLI, Codex CLI and anything else supporting SKILL.md. You never run it on its own.What do I need besides the skill — Node, a specific browser?CDP browser mode needs Node.js 22+ and Chrome or Edge with remote debugging enabled: open chrome://inspect/#remote-debugging (or edge://inspect on Edge) and tick 'Allow remote debugging for this browser instance', restarting the browser if asked.
CDP browser mode needs Node.js 22+ and Chrome or Edge with remote debugging enabled: open chrome://inspect/#remote-debugging (or edge://inspect on Edge) and tick 'Allow remote debugging for this browser instance', restarting the browser if asked.Could the browser automation get my accounts in trouble?Yes — automating social platforms like Xiaohongshu through the browser risks rate-limiting or bans, so the author strongly recommends a burner account. Also, if your preferred browser isn't running with debugging on, the proxy fails hard instead of silently switching.
Yes — automating social platforms like Xiaohongshu through the browser risks rate-limiting or bans, so the author strongly recommends a burner account. Also, if your preferred browser isn't running with debugging on, the proxy fails hard instead of silently switching.Does it work on Windows, or is it Mac and Linux only?All three desktops: since v2.4.1 the scripts run on Node.js instead of bash, so Windows, Linux and macOS are all supported.
All three desktops: since v2.4.1 the scripts run on Node.js instead of bash, so Windows, Linux and macOS are all supported.Why drive my real browser instead of a headless one?The CDP proxy connects to your everyday Chrome or Edge, so it naturally carries your login sessions and handles dynamic pages, real clicks and file uploads that plain HTTP fetches can't. It even grabs video frames from pages for analysis.
The CDP proxy connects to your everyday Chrome or Edge, so it naturally carries your login sessions and handles dynamic pages, real clicks and file uploads that plain HTTP fetches can't. It even grabs video frames from pages for analysis.What's the 'site experience' feature about?The skill stores operational know-how per domain — URL patterns, platform quirks, known pitfalls — and reuses it across sessions, so it keeps getting better at the sites you actually use, including tricky ones like Xiaohongshu.
The skill stores operational know-how per domain — URL patterns, platform quirks, known pitfalls — and reuses it across sessions, so it keeps getting better at the sites you actually use, including tricky ones like Xiaohongshu.Can it research several targets at the same time?Yes — for multi-target tasks it dispatches sub-agents that run in parallel, all sharing one CDP proxy with tab-level isolation between them. A typical ask: research five product websites at once and return a comparison summary.
Yes — for multi-target tasks it dispatches sub-agents that run in parallel, all sharing one CDP proxy with tab-level isolation between them. A typical ask: research five product websites at once and return a comparison summary.Can it find pages I've visited before but can't remember?Yes — the bundled find-url.mjs searches your local Chrome and Edge bookmarks and history by keyword, time window or visit frequency, which is handy for internal systems that public search engines simply can't find.
Yes — the bundled find-url.mjs searches your local Chrome and Edge bookmarks and history by keyword, time window or visit frequency, which is handy for internal systems that public search engines simply can't find.What are similar projects?bb-browser, oc, pstack-claude, nuwa-skill, skills, draw-architecture, skills, agentic-awesome-skills
bb-browser, oc, pstack-claude, nuwa-skill, skills, draw-architecture, skills, agentic-awesome-skillsWhich topics is it in?cdp, codex-cli, chrome, gemini-cli, web-scraping, browser-automation, skill, cursor, browser, developers, agent-skills, claude-code
cdp, codex-cli, chrome, gemini-cli, web-scraping, browser-automation, skill, cursor, browser, developers, agent-skills, claude-code