agent-browser - SKILL.md Agent Skill

name: agent-browser description: Headless browser automation via the agent-browser CLI (Playwright). Use when you need deterministic navigation, DOM interaction, form filling, screenshots/PDFs, or accessibility snapshots with refs for AI-driven selection, especially on JS-heavy pages or scripted browser flows.

Agent Browser

Setup

Install the CLI and download Chromium:

npm install -g agent-browser
agent-browser install

Set a custom browser binary if needed:

AGENT_BROWSER_EXECUTABLE_PATH=/path/to/chromium agent-browser open https://example.com

Quick Start

agent-browser open https://example.com
agent-browser snapshot -i --json
agent-browser click @e2
agent-browser fill @e3 "test@example.com"
agent-browser get text @e1
agent-browser screenshot page.png
agent-browser close

Snapshot + Ref Workflow

Open a page with agent-browser open <url>.
Capture a focused tree with agent-browser snapshot -i -c -d 5 --json.
Choose refs (@e1, @e2, ...) from the snapshot and act with click, fill, press, or hover.
Re-run snapshot after navigation or UI changes.
Close the session with agent-browser close.

Common Commands

Navigate: open, wait --url, get url, get title
Interact: click, dblclick, fill, type, press, hover, scroll
Extract: snapshot, get text, get html, get value, get attr
Output: screenshot --full, pdf <path>
State: cookies, storage local, storage session
Settings: set viewport, set headers <json>, set device, set geo, set offline

Session + Debug

Use isolated sessions with --session <name> or AGENT_BROWSER_SESSION.
Show a visible window with --headed for debugging.
Attach to an existing browser via CDP with --cdp 9222.
Prefer --json for machine-readable output.