메뉴
Browser automation using OpenClaw's built-in browser tool. Use when the user wants to navigate websites, check emails, access web applications, take screenshots, fill forms, or perform any web-based tasks. This skill controls a real Chrome/Chromium browser with persistent profiles.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | openclaw-browser |
| description | Browser automation using OpenClaw's built-in browser tool. Use when the user wants to navigate websites, check emails, access web applications, take screenshots, fill forms, or perform any web-based tasks. This skill controls a real Chrome/Chromium browser with persistent profiles. |
Control a real Chrome/Chromium browser through the OpenClaw CLI to navigate websites, interact with web apps, take screenshots, and extract information.
browser.enabled: true):0.0)Always follow this sequence:
export DISPLAY=:0.0
openclaw browser status
openclaw browser start # If not running
openclaw browser navigate <URL>
openclaw browser open <URL> # Opens in new tab
openclaw browser snapshot # AI-friendly page structure with refs
openclaw browser screenshot # Capture screenshot
openclaw browser screenshot --full-page
openclaw browser tabs # List open tabs
Use refs from snapshot output:
openclaw browser click <ref>
openclaw browser type <ref> "text to type" [--submit]
openclaw browser fill --fields '[{"ref":"1","value":"text"}]'
openclaw browser press Enter
openclaw browser scrollintoview <ref>
openclaw browser select <ref> "Option1" "Option2"
openclaw browser upload /path/to/file
openclaw browser wait --text "Expected text"
openclaw browser wait --selector "css-selector"
sleep 2 # Simple wait for page load
openclaw browser evaluate --fn '(el) => el.textContent' --ref <ref>
openclaw browser evaluate --fn '() => document.title'
export DISPLAY=:0.0
openclaw browser start
openclaw browser navigate https://mail.proton.me
sleep 3
openclaw browser snapshot # Check if logged in
# If login page appears, user must log in manually (persistent profile will remember)
openclaw browser screenshot
# After navigating to inbox
openclaw browser snapshot | grep -A5 "region.*Unread\|region.*message"
# Click on the first email region ref
openclaw browser click <ref>
sleep 2
openclaw browser snapshot # Read content
openclaw browser screenshot
# 1. Start/navigate
export DISPLAY=:0.0
openclaw browser status || openclaw browser start
openclaw browser navigate <URL>
sleep 2
# 2. Get page state
openclaw browser snapshot
openclaw browser screenshot
# 3. Interact using refs from snapshot
openclaw browser click <ref>
openclaw browser type <ref> "value"
# 4. Verify result
openclaw browser screenshot
openclaw profile keeps logins/cookies between sessionssleep 2-3 after navigation for pages to loadsnapshot output to understand page structure and find refsBrowser won't start (display error)
DISPLAY=:0.0 is setps aux | grep XorgElement not found
Login required on every visit
openclaw profile (default)Set up a Microsoft Teams/Microsoft 365 Teams SDK bot that bridges Teams chat to a Nous Research Hermes Agent HTTP API.
Deploy a randomized local Docker Hermes agent with Ollama Cloud and SearXNG.
Deploy OpenClaw gateway instances on Kubernetes with SearXNG as the web search provider. Use when the user wants to (1) Deploy a new OpenClaw instance, (2) Configure or verify SearXNG search integration, (3) Generate OpenClaw configuration files, (4) Set up K8s manifests for OpenClaw + SearXNG, or (5) Troubleshoot search provider connectivity in an OpenClaw deployment.
Manage and inspect Shrubnet's n8n automation server via the n8n Public API. Use when the user asks about n8n workflows, executions, credentials metadata, tags, users, webhook automation, debugging workflow runs, activating/deactivating workflows, creating/updating workflows, or checking the n8n server at n8n.cluster.shrubnet.com. Stores API credentials in the skill folder .env, never in SKILL.md.
Manage DNS records on Cloudflare via the Cloudflare MCP server or Cloudflare API v4. Use this skill whenever the user wants to view, create, update, or delete DNS records for a domain managed by Cloudflare, or when redirecting a domain on Cloudflare to a hosting provider (such as Netlify, DigitalOcean, Vercel, or any origin server). Triggers include any mention of "Cloudflare DNS", "Cloudflare", "CF DNS", "orange cloud", "proxied", "CNAME flattening", or general DNS management terms like "DNS records", "A record", "CNAME", "point domain to" when the domain is known to be on Cloudflare. Also triggers when another skill (like netlify-landing-page-deploy or digitalocean-management) needs DNS records updated on a Cloudflare-managed domain. This skill handles listing zones, reading existing records, creating/updating/deleting individual records, batch operations, and managing Cloudflare-specific features like proxy status (orange/grey cloud). It does NOT handle Cloudflare Workers, Pages, R2, WAF, or other non-DNS
Control the Raspberry Pi (LXDE) desktop via PyAutoGUI MCP. Implements a screenshot-reason-act-verify loop: take a screenshot, analyze what is visible, determine where to click or type, act, then immediately screenshot again to assess the result before proceeding. Use when the user asks to click, type, automate, or interact with any Linux GUI.