{"skill":{"slug":"agent-browser-clawdbot","displayName":"Agent Browser","summary":"Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection","description":"---\nname: agent-browser\ndescription: Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection\nmetadata: {\"clawdbot\":{\"emoji\":\"🌐\",\"requires\":{\"commands\":[\"agent-browser\"]},\"homepage\":\"https://github.com/vercel-labs/agent-browser\"}}\n---\n\n# Agent Browser Skill\n\nFast browser automation using accessibility tree snapshots with refs for deterministic element selection.\n\n## Why Use This Over Built-in Browser Tool\n\n**Use agent-browser when:**\n- Automating multi-step workflows\n- Need deterministic element selection\n- Performance is critical\n- Working with complex SPAs\n- Need session isolation\n\n**Use built-in browser tool when:**\n- Need screenshots/PDFs for analysis\n- Visual inspection required\n- Browser extension integration needed\n\n## Core Workflow\n\n```bash\n# 1. Navigate and snapshot\nagent-browser open https://example.com\nagent-browser snapshot -i --json\n\n# 2. Parse refs from JSON, then interact\nagent-browser click @e2\nagent-browser fill @e3 \"text\"\n\n# 3. Re-snapshot after page changes\nagent-browser snapshot -i --json\n```\n\n## Key Commands\n\n### Navigation\n```bash\nagent-browser open <url>\nagent-browser back | forward | reload | close\n```\n\n### Snapshot (Always use -i --json)\n```bash\nagent-browser snapshot -i --json          # Interactive elements, JSON output\nagent-browser snapshot -i -c -d 5 --json  # + compact, depth limit\nagent-browser snapshot -s \"#main\" -i      # Scope to selector\n```\n\n### Interactions (Ref-based)\n```bash\nagent-browser click @e2\nagent-browser fill @e3 \"text\"\nagent-browser type @e3 \"text\"\nagent-browser hover @e4\nagent-browser check @e5 | uncheck @e5\nagent-browser select @e6 \"value\"\nagent-browser press \"Enter\"\nagent-browser scroll down 500\nagent-browser drag @e7 @e8\n```\n\n### Get Information\n```bash\nagent-browser get text @e1 --json\nagent-browser get html @e2 --json\nagent-browser get value @e3 --json\nagent-browser get attr @e4 \"href\" --json\nagent-browser get title --json\nagent-browser get url --json\nagent-browser get count \".item\" --json\n```\n\n### Check State\n```bash\nagent-browser is visible @e2 --json\nagent-browser is enabled @e3 --json\nagent-browser is checked @e4 --json\n```\n\n### Wait\n```bash\nagent-browser wait @e2                    # Wait for element\nagent-browser wait 1000                   # Wait ms\nagent-browser wait --text \"Welcome\"       # Wait for text\nagent-browser wait --url \"**/dashboard\"   # Wait for URL\nagent-browser wait --load networkidle     # Wait for network\nagent-browser wait --fn \"window.ready === true\"\n```\n\n### Sessions (Isolated Browsers)\n```bash\nagent-browser --session admin open site.com\nagent-browser --session user open site.com\nagent-browser session list\n# Or via env: AGENT_BROWSER_SESSION=admin agent-browser ...\n```\n\n### State Persistence\n```bash\nagent-browser state save auth.json        # Save cookies/storage\nagent-browser state load auth.json        # Load (skip login)\n```\n\n### Screenshots & PDFs\n```bash\nagent-browser screenshot page.png\nagent-browser screenshot --full page.png\nagent-browser pdf page.pdf\n```\n\n### Network Control\n```bash\nagent-browser network route \"**/ads/*\" --abort           # Block\nagent-browser network route \"**/api/*\" --body '{\"x\":1}'  # Mock\nagent-browser network requests --filter api              # View\n```\n\n### Cookies & Storage\n```bash\nagent-browser cookies                     # Get all\nagent-browser cookies set name value\nagent-browser storage local key           # Get localStorage\nagent-browser storage local set key val\n```\n\n### Tabs & Frames\n```bash\nagent-browser tab new https://example.com\nagent-browser tab 2                       # Switch to tab\nagent-browser frame @e5                   # Switch to iframe\nagent-browser frame main                  # Back to main\n```\n\n## Snapshot Output Format\n\n```json\n{\n  \"success\": true,\n  \"data\": {\n    \"snapshot\": \"...\",\n    \"refs\": {\n      \"e1\": {\"role\": \"heading\", \"name\": \"Example Domain\"},\n      \"e2\": {\"role\": \"button\", \"name\": \"Submit\"},\n      \"e3\": {\"role\": \"textbox\", \"name\": \"Email\"}\n    }\n  }\n}\n```\n\n## Best Practices\n\n1. **Always use `-i` flag** - Focus on interactive elements\n2. **Always use `--json`** - Easier to parse\n3. **Wait for stability** - `agent-browser wait --load networkidle`\n4. **Save auth state** - Skip login flows with `state save/load`\n5. **Use sessions** - Isolate different browser contexts\n6. **Use `--headed` for debugging** - See what's happening\n\n## Example: Search and Extract\n\n```bash\nagent-browser open https://www.google.com\nagent-browser snapshot -i --json\n# AI identifies search box @e1\nagent-browser fill @e1 \"AI agents\"\nagent-browser press Enter\nagent-browser wait --load networkidle\nagent-browser snapshot -i --json\n# AI identifies result refs\nagent-browser get text @e3 --json\nagent-browser get attr @e4 \"href\" --json\n```\n\n## Example: Multi-Session Testing\n\n```bash\n# Admin session\nagent-browser --session admin open app.com\nagent-browser --session admin state load admin-auth.json\nagent-browser --session admin snapshot -i --json\n\n# User session (simultaneous)\nagent-browser --session user open app.com\nagent-browser --session user state load user-auth.json\nagent-browser --session user snapshot -i --json\n```\n\n## Installation\n\n```bash\nnpm install -g agent-browser\nagent-browser install                     # Download Chromium\nagent-browser install --with-deps         # Linux: + system deps\n```\n\n## Credits\n\nSkill created by Yossi Elkrief ([@MaTriXy](https://github.com/MaTriXy))\n\nagent-browser CLI by [Vercel Labs](https://github.com/vercel-labs/agent-browser)\n","topics":["Browser Automation"],"tags":{"latest":"0.1.0"},"stats":{"comments":1,"downloads":124322,"installsAllTime":1337,"installsCurrent":911,"stars":428,"versions":1},"createdAt":1769032854381,"updatedAt":1778485848771},"latestVersion":{"version":"0.1.0","createdAt":1769032854381,"changelog":"- Initial release of agent-browser-clawdbot skill.\n- Enables fast, headless browser automation via CLI, optimized for AI agent workflows.\n- Provides accessibility tree snapshots with ref-based deterministic element selection.\n- Supports session isolation, state persistence, and robust automation commands (navigation, interaction, data extraction, network control, and more).\n- Includes example workflows and best practices for efficient automation.","license":null},"metadata":{"setup":[],"os":null,"systems":null},"owner":{"handle":"matrixy","userId":"s17e24j3hd5g35an84aqwke9px88507e","displayName":"MaTriXy","image":"https://avatars.githubusercontent.com/u/3080139?v=4"},"moderation":null}