What roles are included in the default agent roster?

The roster includes specialized roles like Orchestrator, Explorer, Frontend, Backend, API, iOS, Android, Database, Debugger, Copywriter, and Reviewer.

How do I enable agent teams in Claude Code?

Agent teams are experimental. You must enable them by setting 'CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS': '1' in your settings.json file.

Does using multiple agents increase token costs?

Yes, especially in Discussion mode. However, this skill includes strict optimization rules like scoped searching and silent thinking to minimize unnecessary token consumption.

What is the difference between Hub-and-Spoke and Discussion mode?

Hub-and-Spoke routes all communication through an Orchestrator for execution tasks, while Discussion mode allows agents to debate directly to solve complex architectural or debugging problems.

Agent Team Orchestrator

Name: Agent Team Orchestrator
Author: ZoranSpirkovski

byZoranSpirkovski

0•

생산성 및 워크플로

Coordinates specialized AI sub-agents to manage complex, multi-domain development tasks through structured communication and token-efficient delegation.

The Creating Agent Teams skill transforms Claude Code into a sophisticated multi-agent system designed for large-scale software engineering. It introduces a structured roster of specialized agents—including Frontend, Backend, Database, and Reviewer roles—each assigned specific model tiers like Opus or Sonnet based on task complexity. The skill provides two distinct operational modes: Hub-and-Spoke for streamlined feature implementation and Discussion Mode for adversarial debugging and architectural decision-making. By enforcing strict scoping rules and token optimization protocols, it allows Claude to handle complex full-stack projects without losing context or inflating costs.

주요 기능

01Dual communication architectures: Hub-and-Spoke for implementation and Peer-to-Peer for debate

02Automated discussion logging for permanent records of architectural decisions and hypotheses

030 GitHub stars

04Quality control via dedicated Reviewer and Debugger agents for automated validation

05Specialized agent roster with predefined roles, model tiers, and file-system scopes

06Advanced token optimization including silent thinking and selective file reading

사용 사례

01Architectural planning where different agents advocate for specific design patterns before implementation

02Full-stack feature development requiring synchronized changes across UI, API, and database layers

03Deep-dive debugging of intermittent issues using multiple 'Panelist' agents to test competing hypotheses

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zoranspirkovski/skills creating-agent-teams

For use in Claude.ai and ChatGPT

주요 기능

01Dual communication architectures: Hub-and-Spoke for implementation and Peer-to-Peer for debate

02Automated discussion logging for permanent records of architectural decisions and hypotheses

030 GitHub stars

04Quality control via dedicated Reviewer and Debugger agents for automated validation

05Specialized agent roster with predefined roles, model tiers, and file-system scopes

06Advanced token optimization including silent thinking and selective file reading

사용 사례

01Architectural planning where different agents advocate for specific design patterns before implementation

02Full-stack feature development requiring synchronized changes across UI, API, and database layers

03Deep-dive debugging of intermittent issues using multiple 'Panelist' agents to test competing hypotheses