multi-agent-patterns

Supervisor architecture centralizes multi-agent coordination through a managing agent, while swarm architecture relies on decentralized agent handoffs. This Skill guides pattern selection based on your workflow coordination needs and context isolation requirements.

Design context isolation by defining strict communication protocols that pass only necessary state between specialized agents during handoffs. This Skill provides coordination strategies to prevent context bloat and maintain workflow integrity across parallel AI task execution.

Hierarchical multi-agent design is needed when workflows require layered delegation, splitting complex tasks into specialized sub-tasks. This Skill helps architect these layers to overcome single-agent context limits and manage intricate task dependencies.

Handle failure mitigation by designing consensus strategies and fallback mechanisms within your multi-agent architecture. This Skill addresses these coordination challenges, ensuring individual agent failures do not collapse the entire workflow orchestration.

Yes, you can design a production research system where specialized agents independently gather, analyze, verify, and combine findings. This Skill solves the architectural planning needed for coordinated workflows, communication protocols, and context management.