daffy0208 avatar

multi-agent-architect

Design and orchestrate multi-agent systems. Use when building complex AI systems requiring specializ

by daffy0208|Open Source

Multi-Agent Architect - Quick Start

Version: 1.0.0 | Category: AI-Native Development | Difficulty: Advanced

What This Skill Does

Design and orchestrate multi-agent systems where specialized agents collaborate to solve complex tasks through clear communication patterns and coordination strategies.

Quick Start

  1. Choose pattern - Sequential, Hierarchical, Peer, or Swarm
  2. Define agents - Clear roles and responsibilities (2-5 agents)
  3. Communication - Shared memory, message passing, or events
  4. Implement - CrewAI (role-based) or LangGraph (state machines)
  5. Monitor - Track execution, costs, failures

Success Criteria

  • ✅ Agents have clear, non-overlapping responsibilities
  • ✅ Communication pattern chosen and implemented
  • ✅ Workflow handles agent failures gracefully
  • ✅ Execution traced and logged
  • ✅ Cost per workflow tracked

Key Patterns

  • Sequential: A → B → C → D (clear dependencies)
  • Hierarchical: Manager → Workers (parallel subtasks)
  • Peer: Collaboration with review (quality through iteration)
  • Swarm: Explore solutions, pick best

Decision: Single vs Multi-Agent

  • Single: Task is simple, linear, single expertise
  • Multi: Requires specialization, parallelization, or review

License

Part of ai-dev-standards repository.

Download Skill