dev-ccg

Orchestrate tri-model execution across Claude, Codex, and Gemini to synthesize consensus answers.

520|175|Updated Apr 8, 2026

Search Tags:#architecture #code review #orchestration #multi-model #consensus #bias reduction #LLM synthesis

Ready to run

npx skills add https://github.com/evolution-foundation/evo-nexus --skill dev-ccg-evolution-foundation

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: dev-ccg
Source: https://github.com/evolution-foundation/evo-nexus/tree/main/.claude/skills/dev-ccg
Command: npx skills add https://github.com/evolution-foundation/evo-nexus --skill dev-ccg-evolution-foundation

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It reduces single-model bias by running the same task across Claude, Codex, and Gemini in parallel, then synthesizing a consensus answer for higher-confidence outputs.

Core Features & Use Cases

Tri-model orchestration (CCG): Executes one model-agnostic prompt through Claude, Codex, and Gemini, then merges results into a single synthesis.
Agreement vs disagreement labeling: Clearly separates areas of consensus from conflicting claims so you know what to trust vs review.
High-stakes engineering support: Useful for architecture decisions and code review scenarios where adversarial cross-validation matters.
Degrades gracefully: If Codex or Gemini is unavailable, it warns and falls back to single-model mode rather than failing silently.

Quick Start

Ask to run the dev-ccg skill for your specific high-stakes task and provide the prompt you want evaluated across Claude, Codex, and Gemini.

Frequently Asked Questions about dev-ccg

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

FAQPage Schema

How do I reduce single-model bias when making high-stakes architecture decisions?▼

Multi-model consensus orchestration reduces single-model bias by executing the same architecture prompt across Claude, Codex, and Gemini in parallel, then synthesizing a consensus answer and flagging disagreement regions for human review.

What is the best way to cross-validate LLM outputs for adversarial code review?▼

Tri-model orchestration cross-validates LLM outputs for adversarial code review by running your prompt through Claude, Codex, and Gemini, comparing their responses, and labeling areas of agreement versus conflict.

How do I synthesize a consensus answer from multiple LLM responses?▼

You synthesize a consensus answer from multiple LLM responses by collecting the parallel outputs from Claude, Codex, and Gemini, comparing their claims, and merging the agreements into a single output while explicitly flagging conflicting claims.

Does multi-model consensus orchestration still work if Codex or Gemini is unavailable?▼

Multi-model consensus orchestration degrades gracefully if Codex or Gemini is unavailable by issuing a warning and falling back to single-model mode, rather than failing silently or halting the execution pipeline.

When do I need tri-model consensus for software engineering tasks?▼

You need tri-model consensus for software engineering tasks involving high-stakes decisions like architecture tradeoffs, adversarial code review, and complex planning where cross-model agreement increases confidence and highlights disagreement regions.

Can I use multi-model orchestration to compare Claude, Codex, and Gemini on the same prompt?▼

Yes, multi-model orchestration compares Claude, Codex, and Gemini on the same prompt by executing model-agnostic inputs in parallel, collecting the responses, and outputting a synthesis with agreement and disagreement labels.