Multi-Agent System Design Blueprint Prompt

Content

{
  "variables": [
    "task_description"
  ],
  "prompt_text": "Design a multi-agent system to solve the following task: {{task_description}}. Output a blueprint containing: (1) Agent roster — name, role, model recommendation, context window needs; (2) Communication topology — who calls whom, push vs pull; (3) Shared state schema (JSON); (4) Failure modes and fallback strategies; (5) Estimated token cost per full run. Use Mermaid syntax for the topology diagram.",
  "example_output": "Agents: Orchestrator (claude-3-5-sonnet, 8k ctx), Researcher×3 (gpt-4o-mini, 4k ctx), Synthesizer (claude-3-5-sonnet, 32k ctx). Topology: Orchestrator fans out to Researchers, collects results, sends to Synthesizer. Shared state: { task_id, sub_queries[], results[], status }. Failure: if Researcher times out >30s, Orchestrator retries once then continues with partial results.",
  "model_compatibility": [
    "claude-3-5-sonnet",
    "gpt-4o"
  ]
}