# Oh My Gemini CLI: A Context Engineering-Driven Multi-Agent Collaboration Framework

> Oh My Gemini CLI (OmG) is an extended workflow framework designed for Google Gemini CLI, which expands a single conversational assistant into a structured, role-driven multi-agent engineering collaboration system through context engineering.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-03-31T05:44:39.000Z
- 最近活动: 2026-03-31T05:49:36.763Z
- 热度: 152.9
- 关键词: Gemini CLI, 多智能体, 上下文工程, AI工作流, 智能体编排, 软件工程, Claude Code, 大语言模型, 开发工具
- 页面链接: https://www.zingnex.cn/en/forum/thread/oh-my-gemini-cli
- Canonical: https://www.zingnex.cn/forum/thread/oh-my-gemini-cli
- Markdown 来源: floors_fallback

---

## [Introduction] Oh My Gemini CLI: A Context Engineering-Driven Multi-Agent Collaboration Framework

Oh My Gemini CLI (OmG) is an extended workflow framework for Google Gemini CLI. It expands a single conversational assistant into a structured, role-driven multi-agent engineering collaboration system through context engineering. It provides systematic solutions to the pain points of single-conversation AI assistants, supports closed-loop processing of complex tasks, and is an important attempt in the evolution of AI-assisted programming towards engineering.

## Inspiration and Project Background

At the end of 2024, Jeongkyu Shin, CEO of Lablup, proposed that the core competitiveness of Claude Code lies in its overall workflow model rather than the engine itself, and this model works well when applied to Gemini. This observation inspired developer Joonghyun Lee to think about introducing Claude Code's Harness model into Gemini CLI, which led to the birth of the OmG project.

## Core Design Philosophy and Architecture

OmG is a complete context engineering framework, with the core philosophy that complex tasks require a closed-loop process of planning, execution, and review. Its architecture includes four major components:
1. Context Core: Defines system behavior guidelines, workflow conventions, and state management specifications;
2. Agent Roles: Such as Director (coordinator), Planner (planner), etc., each with clear responsibilities;
3. Deep Work Skills: 8 reusable skills (e.g., $plan, $execute);
4. Command Control Layer: Drives workflows via slash commands (e.g., /omg:intent).

## Workflow Mechanism and Model Strategy

**Workflow Closed Loop**: Intent clarification → Workspace preparation → Team formation → Planning and PRD → Execution and verification → Status synchronization (real-time update to state file).
**Model Strategy**: Use Gemini models in layers: gemini-3.1-pro for key decisions, gemini-3-flash for high-load tasks, gemini-3.1-flash-lite for low-risk exploration, balancing quality and cost.

## Core Problems Solved and Latest Features

**Core Problems Solved**:
- Context Confusion: Role separation isolates planning and execution contexts;
- Progress Visibility: Explicit workflow phases and state files;
- Parallel Synchronization: Workspace and taskboard track task status;
- Deep Conversation Interruption: learn-signal hook suppresses automatic prompts;
- Decision-Execution Disconnect: Built-in review and debugging roles.
**Latest Features**: Introduced the $deep-dive skill, enabling a pipeline of trace → deep interview → planning, calculating implementation readiness and capturing hypothetical risks.

## Installation, Usage, and Community Support

**Installation**: Install via the command `gemini extensions install https://github.com/Joonghyun-Lee-Frieren/oh-my-gemini-cli`.
**Verification**: Use /omg:status for a smoke test.
**Community**: Provides multilingual documentation (English, Korean, Japanese, French, Chinese, Spanish), GitHub Actions version checks, GitHub Sponsors support, and an exclusive Landing Page to showcase the concept.

## Conclusion: A New Paradigm for AI-Assisted Programming

OmG represents an important attempt in the evolution of AI-assisted programming towards structure and engineering, proving that large language models can not only serve as question-answering assistants but also become collaborative partners in complex software engineering. Through context engineering and multi-agent orchestration, it provides developers with a reusable, auditable, and scalable AI-driven development methodology.