// Control the local desktop using the `computer` MCP tool from computer-use-mcp. Use when the user asks to operate local Mac/Windows apps, inspect the screen, click UI, type text, press shortcuts, scroll, drag, or interact with native GUI software.
Control native macOS, Windows, and Linux desktop apps through the `open-computer-use` MCP server. Use when the user asks to operate local apps with accessibility-tree context plus screenshots, inspect the screen, click UI, type text, press shortcuts, scroll, drag, or interact with OS-level GUI software.
Control browser pages using the Playwright MCP server. Use when the user asks to open, inspect, navigate, click, type, test, or automate websites, localhost web apps, browser tabs, forms, and web UI.
帮助用户快速配置和使用微信通道功能。当用户想要"配置微信"、"连接微信"、"设置微信机器人"、"weixin setup"、"wechat channel"时使用此技能。
Use CodeScope to analyze any indexed codebase via its graph database (neug) and vector index (zvec). Supports Python, JavaScript/TypeScript, C, and Java (including Hadoop-scale repositories). Covers call graphs, dependency analysis, dead code detection, hotspots, module coupling, architectural layering, commit history, change attribution, semantic code search, impact analysis, full architecture reports, and bug root cause analysis from GitHub issues. Use this skill whenever the user asks about code structure, code dependencies, who calls what, why something changed, finding similar functions, generating architecture reports, understanding module boundaries, analyzing GitHub issues/bugs, finding bug root causes, understanding why a project has many bugs, tracing bugs to code, indexing Java projects, or any question that benefits from a code knowledge graph — even if they don't mention "CodeScope" by name. If a `.codegraph` or similar index directory exists in the workspace, this skill applies.
为开源项目生成专业宣传视频，传入 {owner}/{repo} 参数 Triggers on "生成开源视频", "宣传视频", "oss video", "remotion video".
Build full-stack dashboard applications using React/Next.js + shadcn/ui + Tailwind CSS + Recharts + Express with customizable data sources. Use when creating data visualization dashboards, business intelligence interfaces, monitoring systems, KPI displays, analytics platforms, or any project requiring data visualization with modern UI components and dynamic effects.
| name | computer-use |
| description | Control the local desktop using the `computer` MCP tool from computer-use-mcp. Use when the user asks to operate local Mac/Windows apps, inspect the screen, click UI, type text, press shortcuts, scroll, drag, or interact with native GUI software. |
Use the computer MCP tool from the computer-use MCP server to operate the user's real local desktop.
The only valid tool path for this skill is the MCP tool named computer.
Do not use shell commands to start another desktop automation MCP server, do not install @anthropic-ai/mcp-computer-use-server, and do not edit Qwen settings as a fallback.
If the computer tool is not available in the current tool list, stop and tell the user to restart Qwen Code or reconnect the computer-use MCP server.
For browser pages, websites, localhost web apps, web forms, DOM elements, links, inputs, or browser navigation flows, use the browser-use skill and the Playwright MCP server instead. Use computer-use only when the task requires native OS or app UI that Playwright cannot see.
computer action get_screenshot.get_screenshot.Ask for confirmation before destructive, privacy-sensitive, or externally visible actions, including deleting files, sending messages, submitting forms, making purchases, changing security settings, or entering credentials.
Do not assume the user wants the whole desktop automated. Operate only the app, window, or workflow they asked for.
action: "get_screenshot" to establish the coordinate frame.action: "left_click", action: "right_click", action: "middle_click", action: "double_click", action: "mouse_move", and action: "left_click_drag" for pointer actions.action: "type" for text input.action: "key" for keys or key combinations.action: "scroll" for scrolling.computer; do not look for separate tools named click or screenshot.On macOS, the user may need to grant Accessibility and Screen Recording permissions to the Node/npm process that runs the MCP server.
On Windows, the desktop must be unlocked and interactive for GUI input to work reliably.