# 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.

- Skill: `gabrielmoreira/agent-browser-5` (Agent Skill)
- Install (CLI): `npx skillmds@latest add gabrielmoreira/agent-browser-5`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gabrielmoreira/agent-browser-5/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: gabrielmoreira (https://skillmd.com/u/gabrielmoreira)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/gabrielmoreira/agent-browser-5

---


# 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:

```bash
npm install -g agent-browser && agent-browser install
```

Or via Homebrew: `brew install agent-browser`

## Core Workflow

Every 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 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

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

### Fill a form

```bash
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

```bash
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

```bash
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

1. **Always snapshot before interacting** — refs don't exist until you snapshot
2. **Re-snapshot after navigation** — refs are invalidated when the page changes
3. **Re-snapshot after dynamic changes** — clicking a button that opens a dropdown requires a new snapshot to see the new elements
4. **Use `wait` for slow pages** — `wait --load networkidle` before snapshot on heavy pages
5. **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`.