squad

Orchestrate multi-agent advisory workflows with risk scoring and rubric-based verdicts.

4|3|Updated May 2, 2026
One-click install
npx skills add https://github.com/ggemba/squad-mcp --skill squad-ggemba
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: squad
Source: https://github.com/ggemba/squad-mcp/tree/main/skills/squad
Command: npx skills add https://github.com/ggemba/squad-mcp --skill squad-ggemba

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It reduces risky coding changes by orchestrating a multi-agent advisory workflow that plans, implements (with approval), and consolidates specialist verdicts into a clear go/change/reject outcome.

Core Features & Use Cases

  • Deterministic multi-agent orchestration: classify the work, score risk, select the right specialists, and consolidate their rubric-based decisions into APPROVED/CHANGES_REQUIRED/REJECTED.
  • Two safe modes: implement mode runs the full workflow with user gates, while review mode performs advisory-only feedback on an existing diff without editing files.
  • Task-scoped execution: break a PRD into atomic tasks and run the squad on one task at a time to prevent context bloat.

Quick Start

Use the squad skill to review or implement changes by running /squad:implement --normal add OAuth login flow.

Frequently Asked Questions about squad

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

FAQPage Schema
How do I run a multi-agent code review on an existing Git diff?

To review a Git diff, use review mode to perform advisory-only feedback without editing files, consolidating specialist verdicts into a clear go/change/reject outcome.

What is multi-agent implementation planning for code changes?

Multi-agent implementation planning classifies work, scores risk, selects specialists, and consolidates their rubric-based decisions into APPROVED, CHANGES_REQUIRED, or REJECTED outcomes.

How do I execute a large PRD without causing context bloat?

To execute a large PRD without context bloat, break it into atomic tasks and run task-scoped execution on one task at a time.

Can I automate code implementation with user approval gates?

Yes, you can automate code implementation with user approval gates by using implement mode to run the full multi-agent workflow with strict safety gates.

Does MCP orchestration support deterministic tool execution for risk scoring?

Yes, MCP orchestration supports deterministic tool execution for risk scoring by requiring YAML-frontend discovery and applying strict safety gates during parallel advisory dispatch.

When should I avoid using a multi-agent advisory workflow for code changes?

Avoid using a multi-agent advisory workflow when you need to directly edit files during a review, as review mode is advisory-only and requires implement mode for actual changes.