browser

The browser skill enables natural-language browser automation through the `browse` CLI, allowing users to navigate websites, extract content, interact with web pages, and automate browser-based workflows directly from the command line. It is designed for tasks such as opening URLs, collecting page data, filling forms, clicking interface elements, taking screenshots, and interacting with JavaScript-heavy applications. The skill supports both local browser sessions and remote Browserbase-powered sessions, helping users automate web interactions in environments that may include anti-bot protections, CAPTCHAs, or rate limiting.

The skill provides flexible environment selection between local and remote execution modes. Local mode supports isolated browser sessions, automatic connection to existing Chrome instances, and attachment to specific Chrome DevTools Protocol targets. Remote Browserbase mode adds advanced capabilities such as anti-bot stealth techniques, automatic CAPTCHA solving, residential proxy support, and persistent browser contexts. Commands are available for navigation, browser history management, retrieving page text or HTML, capturing screenshots, and inspecting accessibility snapshots for structured page state analysis. The `browse snapshot` command is emphasized as the preferred method for understanding page structure because it returns accessibility-tree data with reusable element references.

This skill is useful for developers, QA engineers, automation specialists, and AI-assisted workflows that require browser interaction. Common use cases include scraping protected websites, testing web applications, automating repetitive browsing tasks, interacting with authenticated sessions, and working with sites that rely heavily on client-side JavaScript. It is particularly valuable when standard scraping tools fail due to bot detection or when users need reproducible browser automation across local and remote environments.