AI Agent Hub
Back to skills
Browser Act CLI icon

Browser Act CLI

AI Agent Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @user_2cc35c71/browser-act.

About this skill

Problem

When an AI agent must handle login pages, JS-rendered sites, multiple accounts, or sensitive actions, plain HTTP fetching and simple scripts are often insufficient: content depends on rendering, sessions collide, and forms, challenges, proxies, and human approval lack a stable interface. Browser Act packages these browser capabilities into a CLI for agent-driven work.

How It Works

It is not just a fetcher; it is a browser automation entry point for agents:
- Lightweight extraction: fast retrieval of JS-rendered content, positioned as an alternative to WebFetch or curl.
- Session management: multi-browser isolation and multi-account parallelism reduce cross-environment contamination.
- Interaction and evidence: DOM extraction, screenshots, form filling, file upload, and network capture.
- Safety and collaboration: the Confirmation Gate requires explicit approval for browser creation, deletion, and sensitive operations; interactive challenges can be assisted with user authorization; headed mode and remote assist allow human fallback.

Before use, load the CLI guide with get-skills core to read environment state, available browsers, operational directives, and the complete workflow; this output includes safety constraints and should not be skipped or truncated. Boundaries: it fits agent workflows that need browser context but is not a full RPA platform; for sensitive sites, accounts, payments, or compliance data, keep human approval and least privilege.

Use Cases

  • When debugging frontend agents, extract DOM content, take screenshots, and verify form submissions on authenticated dynamic pages.
  • When testing multiple admin accounts, run isolated browser sessions in parallel to avoid Cookie, session, and proxy contamination.
  • When collecting JS-rendered reports, fetch lightweight content first, then open a browser to capture network requests and save screenshots.
  • When a site requires human verification, let the agent start actions and let users finish sensitive steps via headed mode or remote assist.

Best For

  • Engineers who need agents to perform controlled browser actions inside Cursor, Claude Code, or Codex.
  • QA engineers testing multi-account web apps who need session isolation, proxy switching, and screenshot evidence.
  • Product engineers building internal automation workflows that connect DOM extraction, form filling, and approval gates to agents.
  • Platform leads handling site tasks with human approval who need Confirmation Gate and remote assist capabilities.