visual-rpa-skill
neilhexiaoning-alt 作
Web & Frontend Development概要
Visual RPA desktop automation skill.
説明
このスキルでできること
Automates desktop interactions using screen capture and the Qwen vision model (Qwen-VL). It locates UI elements visually without needing DOM access or accessibility APIs, then executes mouse and keyboard actions with screenshot-based verification. Compound instructions are automatically broken into atomic steps.
Works on any desktop UI regardless of whether the app exposes accessibility APIs or a DOM, making it usable on locked-down or legacy applications.
使いどころ
- Send a WeChat message without touching the keyboard
- Open a browser and navigate to a specific URL
- Create a new folder on the desktop via right-click menu
- Type search queries into desktop app input fields
- Automate multi-step app workflows with a single command
ワークフロー例
User asks: open WeChat, find File Transfer Assistant, type hello, and send it
- 1
Capture screen thumbnail to rough-locate the WeChat icon
- 2
Crop full-resolution region around icon for precise coordinates
- 3
Click WeChat icon and verify it opens via screenshot
- 4
Navigate to File Transfer Assistant chat and click input box
- 5
Type hello using clipboard paste and click the send button
3/3 steps succeeded — message sent via WeChat File Transfer Assistant
要件
- DASHSCOPE_API_KEY environment variable
インストール
clawhub install neilhexiaoning-alt/visual-rpa-skillCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。