菜单
Browse websites interactively — open pages, click elements, fill forms, take screenshots, extract data
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Evaluate whether a principal is authorized to request a lookup, action, or disclosure using the principal registry, action classes, group authorization, and outbound policy.
Apply recipient and outbound safety checks before sending email, messages, uploads, webhooks, or remote writes.
Search, fetch, and synthesize public information with citations while treating external inputs as untrusted.
Create, organize, and manage files and structured content in the workspace
Schedule one-time and recurring tasks with optional pre-check scripts
Search the web, cross-reference sources, and produce cited findings
| name | browser-automation |
| description | Browse websites interactively — open pages, click elements, fill forms, take screenshots, extract data |
| version | 1.0.0 |
| tools | ["browser","bash"] |
| triggers | ["browse","open the website","go to","take a screenshot","fill out the form"] |
Use agent-browser to interact with web pages in a real browser environment.
agent-browser open <url> # Open a page
agent-browser snapshot -i # See interactive elements (@e1, @e2, ...)
agent-browser click @e3 # Click an element
agent-browser fill @e5 "text" # Fill a form field
agent-browser screenshot # Take a screenshot
agent-browser pdf # Save page as PDF
agent-browser open <url>agent-browser snapshot -i to see interactive elements@e1, @e2 — use these in commands