browser-automation

name	browser-automation
description	Drive a real browser via Playwright for tasks that need JavaScript execution, login forms, or interactive UI. Use when: scraping a SPA, filling out a form, taking a screenshot, downloading from a JS-protected page, or running an end-to-end check. NOT for: static HTML (use web_fetch), public APIs (use curl), or anything where a simple HTTP GET works.
metadata	{"all_agents":{"emoji":"🌐","requires":{"bins":["python3"]}},"install":"pip install playwright && playwright install chromium"}

Browser Automation

Use Playwright (Python) via run_command. Playwright is not a runtime dependency of all-agents; the user installs it once.

One-time setup

pip install playwright
playwright install chromium

If neither is installed, ask the user to run those commands and then retry.

Headless one-shot

For most tasks, write a tiny inline script. The run_command tool gives you a workspace cwd; place scripts in playwright_scripts/.

# playwright_scripts/fetch_page.py
from playwright.sync_api import sync_playwright
import sys, json

url = sys.argv[1]
with sync_playwright() as p:
    browser = p.chromium.launch(headless=True)
    page = browser.new_page()
    page.goto(url, wait_until="networkidle")
    print(json.dumps({"title": page.title(), "text": page.inner_text("body")[:5000]}))
    browser.close()

Run it:

python playwright_scripts/fetch_page.py "https://example.com"

Common patterns

Screenshot

page.screenshot(path="screenshot.png", full_page=True)

Fill a form

page.goto(url)
page.fill('input[name="email"]', "...")
page.fill('input[name="password"]', "...")
page.click('button[type="submit"]')
page.wait_for_url("**/dashboard")

Wait for content

page.wait_for_selector(".result-card", timeout=10000)

Persistent login (cookies)

Save and reuse storage_state JSON between runs:

ctx = browser.new_context(storage_state="state.json")     # load
ctx.storage_state(path="state.json")                       # save

Decision rules

Situation	Use
Static page, content is in HTML	web_fetch (much faster)
API endpoint exists	run_command curl ...
JS rendering required	this skill
Login wall / cookies	this skill with persistent context
Form submission	this skill

After the run

Stop the browser, summarize what was extracted, and reference any saved files (screenshots, downloaded artifacts) in the workspace.