browse-site

name

browse-site

description

Navigates and interacts with websites using the agent-browser CLI, including clicking elements, filling forms, taking screenshots, extracting page content, and automating multi-step browser workflows. Use when the user asks to open a URL, click something on a page, fill out a web form, take a page screenshot, scrape or extract data from a website, or automate a browser task. Don't use for Playwright, Puppeteer, or Selenium scripts; don't use for browser extension development; don't use for questions about browser APIs, JavaScript, or frontend code that don't require live browser interaction.

Browse Site Skill

Browse and interact with websites using the agent-browser CLI.

Installation Check

Before first use, verify agent-browser is available:

agent-browser --version

If the command is not found, install it:

npm install -g agent-browser
agent-browser install  # Download Chromium

If installation fails, stop and report the error to the user. Do not proceed.

Workflow

1. Navigate — open the target URL:

   agent-browser open <url>
   

   If the command exits with a non-zero code or outputs an error, report it and stop.

2. Snapshot — get interactive elements with stable refs:

   agent-browser snapshot -i -c --json
   

   The JSON output contains an elements array. Each element has a ref field (e.g., "@e2") and a role/name describing it. Use these refs for all subsequent interactions.

3. Interact — act on elements using refs from the latest snapshot:

   agent-browser click @e2
   agent-browser fill @e5 "search term"
   agent-browser type @e7 "typed text"
   

4. Re-snapshot — after any interaction that changes the page, take a new snapshot before acting again. Refs from a previous snapshot are stale after page changes.

5. Repeat until the task is complete, then close the browser:

   agent-browser close
   

Common Commands

Action	Command
Navigate	agent-browser open <url>
Get elements	agent-browser snapshot -i -c --json
Click	agent-browser click @e<n>
Type text	agent-browser type @e<n> "text"
Fill field	agent-browser fill @e<n> "text"
Screenshot	agent-browser screenshot <path>
Get element text	agent-browser get text @e<n>
Get page title	agent-browser get title
Get current URL	agent-browser get url
Wait (milliseconds)	agent-browser wait 2000
Wait (element)	agent-browser wait "#submit-button"
Close browser	agent-browser close

Always provide a path for screenshots (e.g., agent-browser screenshot /tmp/page.png). The command may fail silently without one.

Sessions

Use --session <name> to persist browser state (cookies, local storage) across commands:

agent-browser open https://example.com --session mysite
agent-browser snapshot -i -c --json --session mysite
agent-browser click @e3 --session mysite
agent-browser close --session mysite

Use sessions for login flows so authenticated state is preserved between steps.

Snapshot Options

Option	Description
-i, --interactive	Only interactive elements (buttons, links, inputs)
-c, --compact	Remove empty structural elements
-d, --depth <n>	Limit tree depth
-s, --selector <sel>	Scope to CSS selector
--json	Machine-readable output with refs

Combine options: agent-browser snapshot -i -c -d 5 --json

Error Handling

Situation	Action
agent-browser not found	Install it (see Installation Check), then retry.
open fails (bad URL, network error)	Report the error and URL to the user. Stop.
Snapshot returns empty elements	The page may require a wait; run agent-browser wait 2000 then re-snapshot.
Login blocked by captcha or MFA	Stop and ask the user to complete the manual step, then continue.
Ref not found (stale ref)	Re-snapshot the page; refs change after DOM updates.
Screenshot path missing	Always supply an explicit output path to avoid silent failures.

Tips

• Use --headed on any command to watch the browser visually during debugging: agent-browser open <url> --headed
• Use --session for any workflow that requires login
• Prefer -i -c on snapshots to keep output small and refs easy to read