se-browser-automation

# Browser Automation Skill ## Profile - Use profile: `openclaw` for all automated browsing - Browser is managed by OpenClaw gateway ## Core Patterns ### Navigation - Use `browser navigate` with direct URLs when possible (faster than clicking through menus) - Always `sleep 4-6` after navigation — SPAs load content asynchronously - If page seems blank, try `screenshot` first — content may be rendered but not in DOM snapshot ### Element Interaction - Use `snapshot` to get element refs, then `act` with refs - For iframe content: use `frame` parameter in snapshot - If compact snapshot misses content, try full snapshot or screenshot - GHL and complex SPAs often render in iframes — check both main page and iframe ### Screenshots vs Snapshots - **Snapshot** — DOM tree, good for finding clickable refs - **Screenshot** — Visual image, good for understanding what's actually on screen - When snapshot shows empty/minimal content but screenshot shows a full page → content is in iframe or shadow DOM ### Login Flows 1. Navigate to login page 2. Snapshot to find form fields 3. Click + type credentials 4. Handle 2FA (check email, enter code) 5. Wait for redirect (sleep 4-6 seconds) 6. Verify logged in via screenshot ### Common Issues - **Popups/modals blocking** → Press Escape - **Lazy-loaded content** → Wait longer, refresh, or navigate directly via URL - **iframe content** → Use frame parameter or evaluate JS to find iframe - **evaluate function errors** → No `const`, `let`, or arrow functions. Use `var` and `function(){}` syntax. - **ref required errors** → Must use ref from snapshot, can't use text selectors ### SPA-Specific Tips - SPAs don't fully reload on navigation — URL changes but DOM updates incrementally - Some content only appears after JavaScript execution - If clicking a link doesn't work, try navigating directly to the URL - F5 refresh can help when SPA state gets stuck