mad-teams

Coordinate parallel agent teams for multi-topic research and validation workflows.

Updated May 7, 2026
One-click install
npx skills add https://github.com/tmalcolm-0607/mad-council-claw --skill mad-teams
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mad-teams
Source: https://github.com/tmalcolm-0607/mad-council-claw/tree/main/.claude/skills/mad-teams
Command: npx skills add https://github.com/tmalcolm-0607/mad-council-claw --skill mad-teams

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the chaos of manually coordinating multiple AI agents for complex development workflows. Without standardized team composition templates and dispatch protocols, parallel agent execution leads to file conflicts, duplicated work, and inconsistent synthesis of results.

Core Features & Use Cases

  • Team Templates: Six pre-configured patterns (Research Swarm, Review Board, Parallel Implementation, Parallel Validation, PR Review Board, Debate Panel) for common multi-agent scenarios.
  • Smart Dispatch Protocol: Automated decision logic that evaluates task count, file ownership, and estimated duration to choose between agent teams and sequential subagents.
  • Cost Optimization: Built-in model selection guidance (Sonnet for teammates, Opus for leads) reducing costs by up to 40% while maintaining output quality.
  • Use Case: A development team implementing a complex feature can use this Skill to automatically spawn parallel researchers to investigate multiple technical topics, then coordinate a review board to validate findings, all managed by a lead agent.

Quick Start

Use the mad-teams skill to coordinate a parallel research swarm for investigating authentication, authorization, and session management patterns in the current feature.

Frequently Asked Questions about mad-teams

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I coordinate parallel AI agents without causing file conflicts?

Parallel agent execution requires standardized team templates and dispatch protocols to prevent file conflicts, duplicated work, and inconsistent synthesis. Automated decision logic evaluates task count, file ownership, and estimated duration to manage team composition and lifecycle safely.

When should I use parallel agent teams instead of sequential subagents?

Parallel agent teams are suited for scenarios with three or more independent tasks, disjoint file ownership, and estimated durations exceeding thirty minutes. Automated smart dispatch protocols evaluate these requirements and gracefully fall back to sequential subagent execution when conditions are not met.

What is the best way to optimize costs when running multi-agent research workflows?

Cost optimization in multi-agent workflows uses built-in model selection guidance, assigning Sonnet for teammates and Opus for leads. This hierarchical model selection reduces operational costs by up to 40% while maintaining output quality during parallel execution.

How do I set up a parallel research swarm for multiple technical topics?

Setting up a parallel research swarm involves selecting from six pre-configured team templates like Research Swarm or Review Board. The lead agent automatically spawns parallel researchers to investigate multiple technical topics and coordinates a review board to validate findings.

Does mad-teams support automated task dispatch for complex development workflows?

Automated task dispatch for complex development workflows is supported through smart dispatch protocols. The system applies team dispatch decision rules, manages agent lifecycles, optimizes costs through model selection, and provides graceful fallback to sequential execution.

Can I use pre-configured team templates for parallel validation and PR reviews?

Pre-configured team templates include Parallel Validation and PR Review Board patterns for multi-agent scenarios. These templates provide standardized composition patterns to automate validation and review processes across independent tasks.