When should I use a broadcast message instead of a direct message?

Use broadcast only for critical blockers or major changes to shared resources affecting the entire team, as broadcasting consumes significantly more API resources than targeted direct messages.

Why is using teammate names preferred over agent IDs?

Using names defined in the configuration file ensures messages are human-readable, less error-prone during manual debugging, and properly mapped to the team structure.

How does the plan approval workflow function?

Agents spawned in plan mode must submit a request via the ExitPlanMode tool. The lead agent then reviews the proposed steps and issues an approval or a rejection with specific feedback.

What happens if an agent rejects a shutdown request?

A rejection typically indicates the agent is still completing a critical task. The lead agent should wait for a task completion update before retrying the shutdown request.

Team Communication Protocols

Name: Team Communication Protocols
Author: wshobson

bywshobson

•

29,555

•

协作工具

Establishes structured messaging standards and workflow coordination for multi-agent teams within Claude Code.

This skill provides a standardized framework for communication between autonomous agent teammates, ensuring efficient coordination and resource management. It defines specific protocols for direct messaging, broadcasting critical updates, handling plan approval workflows, and executing graceful team shutdowns. By implementing these protocols, developers can prevent common communication anti-patterns, manage agent idle time effectively, and maintain clear visibility into the collaborative problem-solving process of multi-agent systems.

主要功能

01Structured plan approval workflows for lead-agent oversight

02Automated teammate discovery via configuration file mapping

03Anti-pattern identification to reduce API overhead and noise

0429,555 GitHub stars

05Standardized message type selection for targeted or global communication

06Graceful shutdown and termination sequences to preserve state

使用场景

01Implementing lead-agent-in-the-loop reviews for sensitive code changes

02Managing the lifecycle of transient agent teams to optimize compute costs

03Orchestrating complex software engineering tasks across multiple specialized agents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add wshobson/agents team-communication-protocols

For use in Claude.ai and ChatGPT

Download Skill