菜单
Controls browser instances for web testing, form filling, screenshots, and data extraction. Use when the user asks to browse a website, navigate web pages, fill forms, take screenshots, or test web applications. The agent can start its own browser session or control one started by the user.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Guide for understanding and working with the Oqto platform - a self-hosted AI agent workspace. Use when explaining Oqto architecture, API usage, session management, CLI tools (oqto, oqtoctl, agntz, byt), configuration, or when helping users interact with the platform's features including chat, file browser, terminal, voice mode, and memory system.
Design and implement delightful, scriptable command-line tools: subcommands, composable stdin/stdout/stderr, stable --json/--yaml, great help/errors, safe destructive ops, config precedence (flags>env>config), TTY-aware color/progress, completions, and tests.
This skill should be used to create high quality pdf documents based on predefined templates. When the user asks for a document, you can list all available templates and brands and ask the user for the content to put into the template. You only edit the toml templates, not the .typ files. The toml/typ pair compile into a professional looking pdf via tmpltr.
Fetches official documentation for external libraries and frameworks (React, Next.js, Prisma, FastAPI, Express, Tailwind, MongoDB, etc.) with 60-90% token savings via content-type filtering. Use this skill when implementing features using library APIs, debugging library-specific errors, troubleshooting configuration issues, installing or setting up frameworks, integrating third-party packages, upgrading between library versions, or looking up correct API patterns and best practices. Triggers automatically during coding work - fetch docs before writing library code to get correct patterns, not after guessing wrong.
This skill conducts discovery conversations to understand user intent and agree on approach before taking action. It should be used when the user explicitly calls /interview, asks for recommendations, needs brainstorming, wants to clarify, or when the request could be misunderstood. Prevents building the wrong thing by uncovering WHY behind WHAT.
Create valid code for generating mermaid diagrams
| name | oqto-browser |
| description | Controls browser instances for web testing, form filling, screenshots, and data extraction. Use when the user asks to browse a website, navigate web pages, fill forms, take screenshots, or test web applications. The agent can start its own browser session or control one started by the user. |
| allowed-tools | Bash(oqto-browser:*) |
You can browse the web using oqto-browser. There are two ways to use it:
When you need to browse a website, just use oqto-browser with any session name. The daemon starts automatically:
oqto-browser open https://example.com # Opens browser, navigates to URL
oqto-browser snapshot -i # List interactive elements
oqto-browser click @e1 # Click element
oqto-browser screenshot /tmp/shot.png # Take screenshot
The default session name is default. Use --session <name> for multiple parallel sessions.
When the user starts a browser from the UI and sends instructions to chat, they'll provide a session ID:
oqto-browser --session <SESSION_ID> snapshot -i
oqto-browser --session <SESSION_ID> open <url>
oqto-browser open <url>oqto-browser snapshot -i (returns elements with refs like @e1, @e2)oqto-browser screenshot /tmp/shot.pngoqto-browser open <url> # Navigate to URL
oqto-browser back # Go back
oqto-browser forward # Go forward
oqto-browser reload # Reload page
oqto-browser snapshot # Full accessibility tree
oqto-browser snapshot -i # Interactive elements only (recommended)
oqto-browser click @e1 # Click
oqto-browser fill @e2 "text" # Clear and type
oqto-browser type @e2 "text" # Type without clearing
oqto-browser press Enter # Press key
oqto-browser hover @e1 # Hover
oqto-browser select @e1 "value" # Select dropdown
oqto-browser scroll down 500 # Scroll page
oqto-browser scrollintoview @e1 # Scroll element into view
oqto-browser screenshot /tmp/shot.png # Save screenshot (display with @/tmp/shot.png)
oqto-browser title # Get page title
oqto-browser url # Get current URL
oqto-browser console # View console messages
oqto-browser content # Get page HTML
oqto-browser eval "document.title" # Run JavaScript
oqto-browser wait 2000 # Wait milliseconds
oqto-browser open https://news.ycombinator.com
oqto-browser snapshot -i
# Output: link "Hacker News" [ref=e1], link "new" [ref=e2], ...
oqto-browser screenshot /tmp/hn.png
# Show to user: @/tmp/hn.png
oqto-browser click @e2 # Click "new" link
oqto-browser wait 1000
oqto-browser snapshot -i # See new page
oqto-browser open https://example.com/login
oqto-browser snapshot -i
# Output: textbox "Email" [ref=e1], textbox "Password" [ref=e2], button "Submit" [ref=e3]
oqto-browser fill @e1 "user@example.com"
oqto-browser fill @e2 "password123"
oqto-browser click @e3
oqto-browser wait 2000
oqto-browser snapshot -i # Check result
Add --json for machine-readable output:
oqto-browser snapshot -i --json
oqto-browser close # Close the browser session