multi-agent-patterns

Common multi-agent orchestration failure modes include divergence, error propagation, and consensus hallucination. This Skill catalogs these specific coordination failures and provides mitigation strategies to stabilize agent workflows.

To design a supervisor pattern for coordinating multiple AI agents, this Skill provides specific rules and context isolation principles. It details how a central supervisor manages interactions and distributes tasks across the system.

Token economics impact multi-agent coordination patterns through specific token usage multipliers. This Skill analyzes these costs across Supervisor, Peer-to-Peer, and Hierarchical patterns to help optimize parallel task execution and reduce overhead.

You should use hierarchical agent workflows for structured task delegation and peer-to-peer for direct coordination. This Skill compares these orchestration patterns, highlighting their specific rules and token cost implications for complex system design.

Multi-agent parallel task execution often fails due to context isolation breakdowns or unhandled error propagation. This Skill offers debugging strategies for coordination failures, focusing on context isolation principles and failure mode analysis.