How do I install the Agent Browser skill?

Run npx skillmds@latest add gabrielmoreira/agent-browser-5 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Agent Browser safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: makes network calls, reads secrets. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Agent Browser?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Agent Browser free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Agent Browser?

gabrielmoreira (@gabrielmoreira) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Agent Browser by gabrielmoreira — Agent Skill

Name: Agent Browser
Rating: 4.2 (5 reviews)
Author: gabrielmoreira

Agent Browser

Browser automation skill using the agent-browser CLI for EXTERNAL WEBSITES via Chrome/CDP. Use this skill when the user asks you to interact with websites, fill forms, scrape pages, take screenshots, or test web apps in Chrome. Trigger on phrases like "open this URL", "click the button", "fill the form", "screenshot the page", "scrape the site", "test the login flow", or "browse to". ⚠️ NOT for Clawfirm app UI — use browser-agent (clawfirm-test skill) for that. ⚠️ NOT the same as browser-agent — these are two completely different tools.

gabrielmoreira Updated Sep 21, 2026 17 repo stars

File contents

agent-browser

⚠️ This skill is for external websites via Chrome/CDP. To automate the Clawfirm app UI (WKWebView), use browser-agent instead — see the clawfirm-test skill. agent-browser ≠ browser-agent: different binaries, different protocols, different targets.

A browser automation skill powered by the agent-browser CLI. It controls Chrome/Chromium via CDP and uses compact element refs for fast, token-efficient interactions.

Prerequisites

Install agent-browser if not already available:

npm install -g agent-browser && agent-browser install

Or via Homebrew: brew install agent-browser

Core Workflow

Every browser automation follows this pattern:

Open a URL: agent-browser open <url>
Snapshot interactive elements: agent-browser snapshot -i
Interact using refs from the snapshot: agent-browser click @e1, agent-browser fill @e2 "text"
Re-snapshot after any page change to get fresh refs

Commands persist via a background daemon — chain them with && in a single bash call or run them sequentially.

Quick Examples

Navigate and extract text

agent-browser open https://example.com
agent-browser snapshot -i
agent-browser get text @e1

Fill a form

agent-browser open https://example.com/login
agent-browser snapshot -i
# Identify the form fields from snapshot output, then:
agent-browser fill @e3 "user@example.com"
agent-browser fill @e4 "password123"
agent-browser click @e5
agent-browser wait --load networkidle
agent-browser snapshot -i

Take a screenshot

agent-browser open https://example.com
agent-browser wait --load networkidle
agent-browser screenshot ./capture.png
# Or full page:
agent-browser screenshot --full ./full-page.png

Scrape data

agent-browser open https://example.com
agent-browser snapshot -i
agent-browser get text body

Key Commands

Command	Description
`open <url>`	Navigate to URL
`snapshot -i`	Get interactive elements with @refs
`click @ref`	Click an element
`fill @ref "text"`	Clear field and type text
`type @ref "text"`	Type without clearing
`select @ref "value"`	Select dropdown option
`check @ref` / `uncheck @ref`	Toggle checkbox
`press Enter`	Press a key
`scroll down 500`	Scroll the page
`get text @ref`	Get element text
`get url`	Get current URL
`get title`	Get page title
`screenshot [path]`	Capture screenshot
`wait @ref`	Wait for element
`wait --load networkidle`	Wait for network idle
`wait --text "Success"`	Wait for text to appear
`close`	Close browser session

Important Rules

Always snapshot before interacting — refs don't exist until you snapshot
Re-snapshot after navigation — refs are invalidated when the page changes
Re-snapshot after dynamic changes — clicking a button that opens a dropdown requires a new snapshot to see the new elements
Use wait for slow pages — wait --load networkidle before snapshot on heavy pages
Close sessions when done — agent-browser close to prevent leaked processes

References

For the full command reference, read references/commands.md. For snapshot/ref details, read references/snapshot-refs.md. For authentication patterns, read references/authentication.md. For session management, read references/session-management.md.

---name: agent-browser-5description: Browser automation skill using the agent-browser CLI for EXTERNAL WEBSITES via Chrome/CDP. Use this skill when the user asks you to interact with websites, fill forms, scrape pages, take screenshots, or test web apps in Chrome. Trigger on phrases like "open this URL", "click the button", "fill the form", "screenshot the page", "scrape the site", "test the login flow", or "browse to". ⚠️ NOT for Clawfirm app UI — use browser-agent (clawfirm-test skill) for that. ⚠️ NOT the same as browser-agent — these are two completely different tools.---# agent-browser> ⚠️ **This skill is for external websites via Chrome/CDP.**> To automate the **Clawfirm app UI** (WKWebView), use `browser-agent` instead — see the **clawfirm-test** skill.> `agent-browser` ≠ `browser-agent`: different binaries, different protocols, different targets.A browser automation skill powered by the `agent-browser` CLI. It controls Chrome/Chromium via CDP and uses compact element refs for fast, token-efficient interactions.## PrerequisitesInstall `agent-browser` if not already available:```bashnpm install -g agent-browser && agent-browser install```Or via Homebrew: `brew install agent-browser`## Core WorkflowEvery browser automation follows this pattern:1. **Open** a URL: `agent-browser open <url>`2. **Snapshot** interactive elements: `agent-browser snapshot -i`3. **Interact** using refs from the snapshot: `agent-browser click @e1`, `agent-browser fill @e2 "text"`4. **Re-snapshot** after any page change to get fresh refsCommands persist via a background daemon — chain them with `&&` in a single bash call or run them sequentially.## Quick Examples### Navigate and extract text```bashagent-browser open https://example.comagent-browser snapshot -iagent-browser get text @e1```### Fill a form```bashagent-browser open https://example.com/loginagent-browser snapshot -i# Identify the form fields from snapshot output, then:agent-browser fill @e3 "user@example.com"agent-browser fill @e4 "password123"agent-browser click @e5agent-browser wait --load networkidleagent-browser snapshot -i```### Take a screenshot```bashagent-browser open https://example.comagent-browser wait --load networkidleagent-browser screenshot ./capture.png# Or full page:agent-browser screenshot --full ./full-page.png```### Scrape data```bashagent-browser open https://example.comagent-browser snapshot -iagent-browser get text body```## Key Commands| Command | Description ||---------|-------------|| `open <url>` | Navigate to URL || `snapshot -i` | Get interactive elements with @refs || `click @ref` | Click an element || `fill @ref "text"` | Clear field and type text || `type @ref "text"` | Type without clearing || `select @ref "value"` | Select dropdown option || `check @ref` / `uncheck @ref` | Toggle checkbox || `press Enter` | Press a key || `scroll down 500` | Scroll the page || `get text @ref` | Get element text || `get url` | Get current URL || `get title` | Get page title || `screenshot [path]` | Capture screenshot || `wait @ref` | Wait for element || `wait --load networkidle` | Wait for network idle || `wait --text "Success"` | Wait for text to appear || `close` | Close browser session |## Important Rules1. **Always snapshot before interacting** — refs don't exist until you snapshot2. **Re-snapshot after navigation** — refs are invalidated when the page changes3. **Re-snapshot after dynamic changes** — clicking a button that opens a dropdown requires a new snapshot to see the new elements4. **Use `wait` for slow pages** — `wait --load networkidle` before snapshot on heavy pages5. **Close sessions when done** — `agent-browser close` to prevent leaked processes## ReferencesFor the full command reference, read `references/commands.md`.For snapshot/ref details, read `references/snapshot-refs.md`.For authentication patterns, read `references/authentication.md`.For session management, read `references/session-management.md`.

Frequently asked questions