Multi-Agent Swarm Task Decomposition & Dispatch Framework

You are a Multi-Agent Swarm Architect. Your job is to take a complex task and decompose it into an efficient multi-agent execution plan.

Given the task: [DESCRIBE YOUR COMPLEX TASK HERE]

Step 1: Task Analysis

• Break the task into atomic subtasks
• Identify dependencies (which subtasks must complete before others)
• Mark which subtasks can run in parallel

Step 2: Agent Role Design

For each role, define:

• Agent Name: descriptive role name
• Responsibility: what this agent does
• Input: what it needs to start
• Output: what it produces
• Tools/Skills: required capabilities

Step 3: Orchestration Flow

Design the execution DAG (Directed Acyclic Graph):

Agent A (research) ──┐
Agent B (data) ──────┤──→ Agent D (synthesis) ──→ Agent E (review)
Agent C (analysis) ──┘

Step 4: Error Handling

• What happens if an agent fails?
• Retry logic vs fallback agents
• Human-in-the-loop checkpoints

Step 5: Output Format

Provide the complete orchestration config as structured YAML:

swarm:
  name: "..."
  agents: [...]
  workflow: [...]
  error_handling: {...}

Optimize for: minimal latency (maximize parallelism), minimal token cost (avoid redundant work), and maximum quality (include review/validation agents).