multi-agent-patterns

What problem does it solve?

Single-agent contexts hit limits when tasks are complex or require cross-domain capabilities. This Skill explains how to design multi-agent architectures that distribute work across specialized agents, enabling parallel reasoning and richer tool use. It covers three architectural patterns: Supervisor/Orchestrator, Peer-to-Peer/Swarm, and Hierarchical, plus context isolation, coordination, and failure handling.

Core Features & Use Cases

• Pattern 1: Supervisor/Orchestrator — centralized planning and synthesis
• Pattern 2: Peer-to-Peer/Swarm — distributed handoffs and collaboration
• Pattern 3: Hierarchical — layered strategy, planning, and execution
• Guidelines on context isolation, coordination, and failure modes

Quick Start

Define a supervisor and several specialized workers, implement clear handoff protocols, and run a coordinated task to achieve an objective.