playwright-browser · AI/ML

Browser automation via Playwright. Write scripts, execute via run.js. Write Playwright code to /tmp, execute from skill directory:

node $SKILL_DIR/run.js /tmp/playwright-task.js

For inline code (variables are auto-injected, see below):

node $SKILL_DIR/run.js "const b = await chromium.launch(); const p = await b.newPage(); await p.goto('http://localhost:3000'); console.log(await p.title()); await b.close();"

$SKILL_DIR is where you loaded this file from.

Default: headless (invisible, less intrusive).

Use { headless: false } when user wants to see the browser. You know when that is.

Screenshots to /tmp. Use `slowMo: 100` for debugging.

For inline code, these are available:

BASE_URL - from PLAYWRIGHT_BASE_URL env var
CI_ARGS - browser args for CI (['--no-sandbox', '--disable-setuid-sandbox'])
EXTRA_HEADERS - from PW_HEADER_NAME/VALUE or PW_EXTRA_HEADERS
chromium, firefox, webkit, devices - from playwright

Example:

node $SKILL_DIR/run.js "
const browser = await chromium.launch({ args: CI_ARGS });
const page = await browser.newPage();
await page.goto(BASE_URL || 'http://localhost:3000');
console.log(await page.title());
await browser.close();
"

run.js auto-installs Playwright on first use. No manual setup needed.

For network mocking, auth persistence, multi-tab, downloads, video, traces: [API_REFERENCE.md](API_REFERENCE.md)