cmaven

작업노트

:bulb: Installation, commands, and practical usage of the oh-my-claudecode (omc) plugin, which automatically dispatches complex Claude Code tasks to multiple specialist agents.

[01] What Is omc?

oh-my-claudecode (omc) is a Multi-Agent orchestration framework for Claude Code. It automatically dispatches complex tasks to specialized agents and works out of the box with zero configuration.

Feature Description
No setup needed Works immediately with defaults
Team-first Plan → Design → Execute → Verify pipeline
Natural-language commands No need to memorize commands; instruct in natural language
Auto-parallelization Automatically distributes complex work across agents
19 specialized agents Architecture, research, testing, data science, etc.
Smart routing Haiku for simple tasks, Opus for complex reasoning

[02] Installation

Enter the following commands in the Claude Code prompt in order.



1
2



/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecode

2-2. npm CLI



1



npm i -g oh-my-claude-sisyphus@latest

:warning: The repository is oh-my-claudecode, but the npm package name is oh-my-claude-sisyphus.

2-3. Initial Setup



1
2



/setup
/omc-setup

Or in the terminal:



1



omc setup

2-4. Quick Start

You can use it immediately after installation:



1



autopilot: build a REST API for managing tasks

[03] Main Commands

3-1. Full Command Map

graph TD
    OMC["omc"]

    OMC --> TEAM["/team\nTeam orchestration"]
    OMC --> AUTO["/autopilot\nFully autonomous execution"]
    OMC --> RALPH["/ralph\nPersistence mode"]
    OMC --> ULW["/ultrawork\nMax parallelization"]
    OMC --> DI["/deep-interview\nRequirements analysis"]
    OMC --> ASK["/ask\nExpert consultation"]
    OMC --> CCG["/ccg\nTriple-model synthesis"]
    OMC --> SKILL["/skill\nSkill management"]

    style OMC fill:#f3e5f5,stroke:#6a1b9a
    style TEAM fill:#e3f2fd,stroke:#1565c0
    style AUTO fill:#e8f5e9,stroke:#2e7d32
    style RALPH fill:#fff3e0,stroke:#e65100
    style ULW fill:#fce4ec,stroke:#c62828
    style DI fill:#f5f5f5,stroke:#616161
    style ASK fill:#ffffcc,stroke:#f9a825
    style CCG fill:#e0f7fa,stroke:#00838f
    style SKILL fill:#fce4ec,stroke:#ad1457

3-2. Command Summary

Command Description When to use
/team Team-based pipeline (plan→design→execute→verify) Structured development work
/autopilot Fully autonomous execution Fast implementation of clear requirements
/ralph Retry until completion Stubborn refactoring jobs
/ultrawork (/ulw) Max parallelization Bulk error fixes
/deep-interview Socratic requirements analysis Clarifying ambiguous requirements
/ask Expert consultation with a specific model Design review, risk analysis
/ccg Codex + Gemini + Claude synthesis Multi-perspective review
/skill Skill management (add/remove/search) Registering recurring patterns

[04] Team Mode — Team Orchestration

This is omc’s core feature. It handles work in a plan → design → execute → verify → revise pipeline.



1



/team 3:executor "fix all TypeScript errors"
Syntax Meaning
3:executor Deploy 3 executor agents
"fix all ..." Task instruction

Usage from tmux CLI



1
2
3



omc team 2:codex "review auth module for security issues"
omc team 2:gemini "redesign UI components for accessibility"
omc team 1:claude "implement the payment flow"

Status check and shutdown



1
2



omc team status auth-review      # Check status
omc team shutdown auth-review    # Shut down

[05] Autopilot — Fully Autonomous Execution

When requirements are clear, run the whole thing autonomously from start to finish.



1



/autopilot "build a todo app"

Or natural-language shortcut:



1



autopilot: build a REST API for managing tasks

[06] Ralph — Persistence Mode

Doesn’t give up and keeps trying until completion. Good for complex refactors or migrations.



1



/ralph "refactor authentication module"

[07] Deep Interview — Requirements Analysis

Use when requirements are unclear, or you want to scrutinize the design first.



1



/deep-interview "I want to build a task management app"

Through Socratic questioning, it:

  • Surfaces hidden assumptions
  • Measures clarity across multiple aspects
  • Aims to know exactly what to build before writing any code

[08] Ultrawork — Max Parallelization

Process bulk work with maximum parallelism.



1



/ultrawork "fix all errors"

Short form:



1



/ulw "fix all errors"

[09] Ask & CCG — Expert Consultation

9-1. Ask — Single-Model Consultation

Request expert advice from a specific model.



1
2
3



/ask claude "review this migration plan"
/ask codex "identify architecture risks"
/ask gemini "propose UI design ideas"

Terminal CLI:



1
2
3



omc ask claude "review this migration plan"
omc ask codex --prompt "identify architecture risks"
omc ask gemini --prompt "propose UI ideas"

9-2. CCG — Triple-Model Synthesis

Combine the perspectives of Codex + Gemini + Claude.



1



/ccg "review this PR - architecture (Codex) and UI (Gemini)"

[10] Skill System

omc lets you save recurring patterns as Skills for reuse.

10-1. Skill Storage Locations

Scope Path Sharing Priority
Project .omc/skills/ Whole team (versioned) High
User ~/.omc/skills/ All projects Low

10-2. Skill File Example



1
2
3
4
5
6
7
8



# .omc/skills/fix-proxy-crash.md
---
name: Fix Proxy Crash
description: aiohttp proxy crashes on ClientDisconnectedError
triggers: ["proxy", "aiohttp", "disconnected"]
source: extracted
---
Wrap handler at server.py:42 in try/except ClientDisconnectedError...

10-3. Skill Management Commands



1
2
3
4
5
6



/skill list              # List skills
/skill add               # Add a new skill
/skill remove            # Remove a skill
/skill edit              # Edit a skill
/skill search            # Search skills
/learner                 # Auto-extract patterns

[11] CLI Utilities

11-1. Terminal Commands

Command Description
omc setup Initial setup
omc hud Real-time HUD status display
omc doctor Diagnose and clear cache
omc wait Check rate-limit status
omc wait --start Enable auto-resume
omc wait --stop Disable auto-resume

11-2. Autoresearch — Automated Research



1
2
3
4



omc autoresearch
omc autoresearch --mission "improve startup performance" \
  --eval "npm test -- --run src/tests/perf.test.ts"
omc autoresearch init --topic "benchmark onboarding flow"

11-3. Notifications (Telegram/Discord/Slack)

Get notified when work completes.



1
2
3
4
5
6
7
8
9
10
11



# Telegram
omc config-stop-callback telegram --enable \
  --token <bot_token> --chat <chat_id>

# Discord
omc config-stop-callback discord --enable \
  --webhook <url>

# Slack
omc config-stop-callback slack --enable \
  --webhook <url>

[12] In-Session Natural-Language Shortcuts

You can use natural language inside a Claude Code session instead of slash commands.

Natural language Action
autopilot: build me a todo app Run Autopilot
ralph: refactor the auth module Run Ralph
deepsearch for auth middleware Search the codebase
ultrathink about this design Deep reasoning
cancelomc / stopomc Abort execution

[13] Updating



1
2
3
4
5
6
7
8
9



# npm
npm i -g oh-my-claude-sisyphus@latest

# Plugin
/plugin marketplace update omc
/setup

# When something is wrong
/omc-doctor

[14] Summary

Use Case Command Description
Structured development /team 3:executor "task" Team pipeline
Fast implementation /autopilot "task" Fully autonomous
Persistent work /ralph "task" Repeat until done
Bulk processing /ulw "task" Max parallelization
Requirements clarification /deep-interview "idea" Socratic analysis
Expert review /ask codex "review request" Single-model consultation
Multi-angle review /ccg "review request" Triple-model synthesis

Refs: oh-my-claudecode GitHub · Official Docs

참고