Multi-Agent Coordination Instrumentation

概要

The Multi-Agent Coordination skill provides a standardized framework for tracing and observing complex AI agent systems within Claude Code. It enables developers to map out intricate parent-child relationships, track task handoffs between supervisor and worker agents, and visualize the complete flow of work across distributed agent architectures. By implementing consistent span attributes and context propagation patterns, it ensures that every step—from initial planning to final synthesis—is transparent and measurable across popular frameworks like LangGraph, CrewAI, and AutoGen, making it essential for debugging and optimizing autonomous agent teams.

主な機能

• Standardized span attributes for agent identity, framework, and workflow state
• Context propagation management to eliminate orphaned traces in complex runs
• Hierarchical trace mapping for parent-child agent relationships
• Detailed handoff tracking with reason-based logging and task summaries
• Pre-built instrumentation patterns for LangGraph, CrewAI, and AutoGen
• 0 GitHub stars

ユースケース

• Debugging complex supervisor-worker architectures and task delegation
• Auditing multi-agent system decision-making paths for compliance and safety
• Optimizing agent handoff logic to reduce workflow latency and errors