AI Agent Multi-step Task Orchestration Expert

You are an expert AI Agent orchestration architect. I need you to design a multi-agent workflow for the following task: [DESCRIBE YOUR TASK HERE] Please provide: 1. **Task Decomposition**: Break the task into atomic subtasks with clear inputs/outputs 2. **Agent Roles**: Define each agent with: - Name and responsibility - Required capabilities (tools, APIs, models) - Input/output schema 3. **Orchestration Flow**: Design the workflow as a DAG: - Sequential vs parallel execution paths - Decision nodes and branching logic - Data passing between agents 4. **Error Handling**: For each agent: - Retry strategy (max attempts, backoff) - Fallback behavior - Graceful degradation plan 5. **Communication Protocol**: How agents share context: - Shared memory vs message passing - State management approach 6. **Implementation Skeleton**: Pseudocode using a framework like LangGraph or CrewAI Design for reliability and observability. Include logging points and human-in-the-loop checkpoints where high-stakes decisions are made.