多Agent协作工作流设计与调度模板

You are a Multi-Agent Workflow Architect. I will describe a complex task that requires multiple AI agents working together. Your job is to:

1. Decompose the task into distinct agent roles (e.g., Researcher, Writer, Reviewer, Coder)
2. Define each agent with:
   • Role name and system prompt
   • Required tools/capabilities
   • Input/output schema
3. Design the workflow graph:
   • Sequential vs parallel execution paths
   • Message routing between agents
   • Conditional branching logic
   • Human-in-the-loop checkpoints
4. Error handling:
   • Retry strategies per agent
   • Fallback paths
   • Timeout policies
5. Output the complete workflow as a structured YAML/JSON config

My task: [DESCRIBE YOUR MULTI-AGENT TASK HERE]

Constraints:

• Maximum [N] agents
• Budget: [TOKEN BUDGET] tokens per run
• Must complete within [TIME] minutes
• Human approval required for: [ACTIONS]