ccg

name	ccg
description	Claude-Codex-Gemini tri-model orchestration — decompose, query, synthesize
tags	["multi-ai","research"]

When to Use

• Complex decisions needing multiple AI perspectives
• Architecture decisions, design trade-offs, or contentious technical choices
• Cross-validation where different models may disagree
• User says "ccg", "tri-model", or wants multi-perspective analysis

Workflow

Delegates to @omg:ccg agent, which executes the 3-phase protocol:

1. Decompose: Split the request into specialized prompts (Codex: architecture/security, Gemini: UX/alternatives)
2. Query: Run advisors in parallel via CLI:
   • codex exec "<prompt>" --ephemeral -o /dev/stdout
   • gemini -p "<prompt>"
3. Synthesize: Unify all perspectives (Codex + Gemini + Claude's own) with attribution and action checklist

Prerequisites

Full tri-model requires Codex CLI and Gemini CLI. Falls back to available models if one is missing.

• Codex: npm install -g @openai/codex
• Gemini: npm install -g @google/gemini-cli

Trigger Keywords

ccg, tri-model, three perspectives

Example

@omg:ccg Should I use a message queue or direct HTTP between microservices?

Quality Contract

• Decomposed prompts tailored to each model's strengths
• All perspectives synthesized with attribution
• Action checklist in output