Computer Use Automation
高级development最低上下文：64K
Designs and drives computer-use agents that operate a desktop or browser directly via screenshots and UI actions (click, type, scroll, navigate). Plans multi-step GUI workflows, handles verification and recovery between steps, and structures guardrails for safe autonomous execution. Built for models exposing computer use as a native client-side tool.
使用场景
- Automating repetitive web form submissions across internal tools
- Driving a browser to gather and reconcile data from multiple dashboards
- Building an agent that navigates a legacy desktop app without an API
- Running end-to-end UI flows for QA and regression checks
- Orchestrating high-throughput agentic search and document processing
示例提示词
Design a computer-use agent that reconciles invoices between a web ERP and a spreadsheet. The agent should: 1. Open the ERP dashboard and locate the pending invoices view 2. For each invoice, read the amount and vendor from the screen 3. Cross-check the value against the corresponding spreadsheet row 4. Flag mismatches in a report and never modify records without confirmation Provide: - The action loop (observe screenshot -> plan -> act -> verify) - Guardrails for destructive or irreversible actions - Recovery logic when an expected UI element is missing - A stopping condition and human-in-the-loop checkpoints
推荐模型
兼容工具
gemini-cliclaude-codekiroany
模态
输入: text, image
→输出: text, code
相关 Skills
作者
OpenModels Community