Browser Automation

cd skills/browser-automation
npm install playwright
npx playwright install chromium

1) Validate browser stack

node scripts/launch_browser.js ./artifacts

2) Run login/session capture

Edit config/login.example.json and run:

node scripts/login_flow.js config/login.example.json

3) Run a task flow

Edit config/task.example.json and run:

node scripts/run_task.js config/task.example.json

Input contract

• Required:
  • Target URL(s)
  • Task intent (login / form / scrape / screenshot)
• Optional:
  • Selector list and fallback selectors
  • headless mode toggle
  • storageStatePath
  • output directory

Output contract

Return structured JSON including:

• ok
• data (scraped values)
• screenshots (local file paths)
• finalUrl
• error with failing step details (if failed)

Error handling standards

• On timeout: capture failure screenshot and return actionable selector/url details.
• On missing element: include selector attempted and step type.
• On OAuth/cookie expiry: re-run login flow and refresh storage state.
• On captcha/MFA: pause automation and request human intervention.

Edge cases

• Dynamic DOM updates: prefer role/text selectors and explicit waits.
• Redirect chains: use waitForURL with expected pattern.
• Session loss: re-create storageState from login flow.
• Anti-bot challenge: switch to headed mode + manual assist.