多Agent协作工作空间设计与任务编排提示词

You are an expert in multi-agent systems architecture. Help me design a collaborative agent workspace where multiple AI agents work together as a team. ## Workspace Requirements - Team size: [e.g., 3-5 agents with different specializations] - Task types: [e.g., coding, code review, testing, documentation] - Coordination model: [e.g., hierarchical with a lead agent / peer-to-peer] ## Design the following: ### 1. Agent Role Definitions For each agent role, specify capabilities and skill tree, tools/MCP servers it can access, decision authority boundaries, and escalation triggers. ### 2. Task Orchestration - How tasks are decomposed and assigned - Dependency graph management - Parallel vs sequential execution strategy - Progress tracking mechanism ### 3. Communication Protocol - Message format between agents - Shared context/memory design - Conflict resolution when agents disagree - Human-in-the-loop checkpoints ### 4. Skill Evolution - How agents learn from completed tasks - Skill tree growth mechanism - Performance metrics per agent - When to spawn new specialist agents ### 5. Implementation Blueprint - Recommended framework (OpenAI Agents SDK / CrewAI / LangGraph / Multica) - File structure and configuration - Monitoring and observability setup Provide concrete YAML/JSON configurations and Mermaid diagrams where helpful.