# Ohmg

> Ultimate multi-agent framework for Google Antigravity. Orchestrates specialized domain agents (PM, Frontend, Backend, Mobile, QA, Debug) via Serena Memory.

- Skill: `akillness/ohmg` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add akillness/ohmg`
- Raw SKILL.md: https://api.skillmd.com/api/skills/akillness/ohmg/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Web & Frontend
- Author: akillness (https://skillmd.com/u/akillness)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/akillness/ohmg

---



# oh-my-ag (ohmg) - Multi-Agent Orchestrator

## When to use this skill

- Coordinating complex multi-domain projects
- Parallelizing tasks across multiple AI agents (PM, Frontend, Backend, etc.)
- Using Serena Memory for cross-agent state management
- Setting up real-time observability dashboards for agent workflows
- Integrating multi-CLI vendors (Gemini, Claude, Codex) in a single project

---

## Quick Start

**Scenario: Platform setup for a Unity3D project**

```bash
# Step 1: Install oh-my-ag
bunx oh-my-ag
# Select "Unity3D" when prompted for project type
```

```bash
# Step 2: Verify system is ready
bunx oh-my-ag doctor
```

```bash
# Step 3: Launch the monitoring dashboard and spawn your first agent
bunx oh-my-ag dashboard &
oh-my-ag agent:spawn backend "Set up Unity3D build pipeline" session-unity-01
```

---

## Instructions

1. Install `oh-my-ag`, then confirm `doctor` and dashboard checks pass in the Unity workspace.
2. Map each domain to the CLI that should own it before spawning long-lived agents.
3. Keep Serena Memory and monitoring enabled so multi-agent Unity work remains auditable.

## Examples

```text
/coordinate
```

```bash
oh-my-ag agent:spawn qa "triage failing Unity regression tests" session-unity-qa
```

---

## 1. Core Concepts

### Specialized Agents
| Agent | Specialization | Triggers |
|-------|---------------|----------|
| **Workflow Guide** | Coordinates complex projects | "multi-domain", "complex project" |
| **PM Agent** | Requirements, task decomposition | "plan", "break down" |
| **Frontend Agent** | React/Next.js, styling | "UI", "component", "styling" |
| **Backend Agent** | API, database, auth | "API", "database", "auth" |
| **Debug Agent** | Bug diagnosis, RCA | "bug", "error", "crash" |

### Serena Memory
Orchestrator writes structured state to `.serena/memories/` for real-time monitoring and cross-agent coordination.

---

## 2. Installation & Setup

### Prerequisites
- **Bun** (CLI and dashboards)
- **uv** (Serena setup)

### Interactive Setup
```bash
bunx oh-my-ag
```
Select project type to install relevant skills to `.agent/skills/`.

### Verification
```bash
bunx oh-my-ag doctor
```

---

## 3. Usage Patterns

### Explicit Coordination
```text
/coordinate
```
PM planning → agent spawning → QA review.

### Spawning Agents via CLI
```bash
# Spawn backend agent for a specific task
oh-my-ag agent:spawn backend "Implement auth API" session-01
```

### Dashboard Monitoring
- Terminal: `bunx oh-my-ag dashboard`
- Web: `bunx oh-my-ag dashboard:web` (http://localhost:9847)

---

## 4. MCP Connection & Bridging

### SSE Mode (Shared Server)
If environment needs stdio-to-http bridging:
```bash
bunx oh-my-ag bridge http://localhost:12341/sse
```

---

## 5. Configuration

Configure per-agent CLI mapping in `.agent/config/user-preferences.yaml`:
```yaml
agent_cli_mapping:
  frontend: gemini
  backend: codex
  pm: claude
  qa: claude
```

---

## Quick Reference

| Command | Action |
|---------|--------|
| `bunx oh-my-ag` | Interactive installer |
| `/setup` | Agent-side configuration |
| `bunx oh-my-ag doctor` | System check & repair |
| `bunx oh-my-ag update` | Update skills |
| `bunx oh-my-ag usage` | Show quota usage |