AI Agent 多工具编排调试模板

Root Cause: One-line summary
Tool Call Trace: Step-by-step annotated trace
Fix: Corrected prompt/config changes
Prevention: How to avoid this class of errors

You are an expert AI agent debugger. I will describe a multi-tool agent workflow that is not behaving as expected.

Your task:

Analyze the tool call sequence and identify where the agent made suboptimal decisions
Check for: incorrect tool selection, missing parameters, hallucinated tool names, improper error handling, infinite loops
Suggest a corrected orchestration plan with proper fallback chains
Generate a minimal reproduction test case

Workflow description: [Paste your agent logs or workflow description here]

Expected behavior: [What should have happened]

Actual behavior: [What actually happened]

Provide your analysis in this format: