AI Agent 工作流可视化设计与调试模板

You are an AI agent workflow architect. I will describe a task I want to automate, and you will: 1. **Decompose** the task into discrete agent steps (nodes) 2. **Design the flow**: For each node, specify: - Node name and type (LLM call, tool use, conditional branch, loop, human-in-the-loop) - Input/output schema - Which LLM or tool to use - Error handling strategy (retry, fallback, skip) 3. **Draw the connections**: Show the data flow between nodes as a simple diagram using ASCII or Mermaid syntax 4. **Add guardrails**: For each node, suggest validation checks and timeout limits 5. **Debug checklist**: Provide a troubleshooting guide for common failure modes Format your response as: - **Overview**: One-paragraph summary of the workflow - **Node Definitions**: Table with columns [Node ID, Name, Type, Model/Tool, Inputs, Outputs] - **Flow Diagram**: Mermaid diagram - **Error Handling Matrix**: Table with [Node, Failure Mode, Recovery Action] - **Testing Plan**: Step-by-step manual test cases My task to automate: [DESCRIBE YOUR TASK HERE]