React Grab

Playwright automation with your real browser cookies. Pages persist across executions. Output is always JSON: {ok, result, error, url, title, page}

Note: If MCP is available in your environment, prefer using the browser_snapshot and browser_execute MCP tools instead of this skill for better performance.

Usage

grab browser execute "<code>"

Performance Tips

1. Batch multiple actions in a single execute call (3-5x faster)
2. Use maxDepth to limit tree depth: getSnapshot({maxDepth: 5})

# SLOW: 3 separate round-trips
execute "await page.goto('https://example.com')"
execute "await getRef('e1').click()"
execute "return await getSnapshot()"

# FAST: 1 round-trip
execute "await page.goto('...'); await getRef('e1').click(); return await getSnapshot();"

Helpers

• page - Playwright Page object
• getSnapshot(opts?) - Get ARIA tree with refs (e1, e2...). Options: maxDepth
• getRef(id) - Get element by ref ID (chainable). E.g. await getRef('e1').click()
• getRef(id).source() - Get React component source: { filePath, lineNumber, componentName }
• fill(id, text) - Clear and fill input
• drag({from, to, dataTransfer?}) - Drag with custom MIME types
• dispatch({target, event, dataTransfer?, detail?}) - Dispatch custom events
• waitFor(target, opts?) - Wait for selector/ref/state. E.g. waitFor('e1'), waitFor('networkidle')

Common Patterns

execute "await getRef('e1').click()"
execute "await fill('e1', 'hello')"
execute "await waitFor('e1')"
execute "await waitFor('networkidle')"
execute "return await getRef('e1').getAttribute('data-id')"
execute "return await getRef('e1').source()"
execute "return await getSnapshot()"
execute "return await getRef('e1').screenshot()"
execute "return await page.screenshot()"

Multi-Page Sessions

execute "await page.goto('https://github.com')" --page github
execute "return await getSnapshot()" --page github

Docs

Playwright API: https://playwright.dev/docs/api/class-page