web-browser
Allows to interact with web pages by performing actions such as clicking buttons, filling out forms, and navigating links. It works by remote controlling Google Chrome or Chromium browsers using the Chrome DevTools Protocol (CDP). When Claude needs to browse the web, it can use this skill to do so.
Security Assessment
About web-browser
The web-browser skill enables users to interact with web pages by controlling a browser remotely using the Chrome DevTools Protocol (CDP). It allows a programmatic interface to perform actions such as clicking buttons, navigating links, and filling out forms in web pages. This skill is particularly useful for automated web browsing tasks, site exploration, and scraping, without requiring direct interaction with the browser UI. It leverages the power of Google Chrome or Chromium for performing these actions, ensuring high compatibility with modern websites and web technologies.
Key features of this skill include the ability to launch Chrome with remote debugging enabled, navigate between pages, execute JavaScript in the browser's context, take screenshots, and interactively select elements on the page. It also has built-in capabilities to dismiss cookie consent dialogs automatically and log network and console activities in real time. These functionalities are ideal for web automation, testing environments, or when there’s a need to programmatically monitor or control web page interactions. This skill can be used by developers, testers, and automation engineers to simplify web-based tasks and improve efficiency in navigating and interacting with web content.
Target users of this skill include developers working on web automation projects, testers needing to automate browser interactions for testing, and data engineers looking to scrape data or monitor web pages remotely. It is useful in scenarios where interacting with the browser in an automated, controlled, and efficient manner is necessary, such as in testing suites, web scraping bots, and site monitoring scripts.
FAQ
How do I start Chrome for remote control?
You can start Chrome for remote debugging by running `./scripts/start.js` for a fresh profile or `./scripts/start.js --profile` to use an existing profile with cookies and logins.
Can I navigate to a new page in a new tab?
Yes, to open a new tab and navigate to a page, you can use `./scripts/nav.js https://example.com --new`.
What happens if I run the dismiss cookies script?
The `./scripts/dismiss-cookies.js` script automatically accepts or rejects cookie consent dialogs on the page, depending on the options passed (default is accept).
Is there a way to interactively pick elements on a page?
Yes, you can use `./scripts/pick.js` to interactively select elements on the page. You can click to select elements, use Cmd/Ctrl+Click for multi-select, and press Enter to finish.
How are the logs generated and where are they stored?
The logs are automatically created by `start.js` and saved in JSONL format under `~/.cache/agent-web/logs/YYYY-MM-DD/`. You can tail these logs using `./scripts/logs-tail.js`.
Install web-browser
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
mitsuhiko/agent-stuff