multi-agent-workflow

• [Architecture](#architecture)
• [What Is This?](#what-is-this)
• [Quick Start](#quick-start)
• [How It Works](#how-it-works)
• [Real-time Dashboards](#real-time-dashboards)
• [Project Structure](#project-structure)
• [Skill Architecture](#skill-architecture)
• [Skills Overview](#skills-overview)
• [CLI Commands](#cli-commands)
• [Troubleshooting](#troubleshooting)
• [Central Registry](#central-registry-for-multi-repo-setup)
• [License](#license)

```mermaid

flowchart TD

subgraph Workflows["Workflows"]

direction TB

W1["/coordinate"]

W2["/orchestrate"]

W3["/plan"]

W4["/review"]

W5["/debug"]

end

subgraph Orchestration["Orchestration"]

direction TB

PM[pm-agent]

WF[workflow-guide]

ORC[orchestrator]

end

subgraph Domain["Domain Agents"]

direction TB

FE[frontend-agent]

BE[backend-agent]

MB[mobile-agent]

end

subgraph Quality["Quality"]

direction TB

QA[qa-agent]

DBG[debug-agent]

end

Workflows --> Orchestration

Orchestration --> Domain

Domain --> Quality

Quality --> CMT([commit])

```

A collection of Antigravity Skills enabling collaborative multi-agent development. Work is distributed across expert agents:

| Agent | Specialization |

|-------|---------------|

| Workflow Guide | Coordinates complex multi-agent projects |

| PM Agent | Requirements analysis, task decomposition, architecture |

| Frontend Agent | React/Next.js, TypeScript, Tailwind CSS |

| Backend Agent | FastAPI, PostgreSQL, JWT authentication |

| Mobile Agent | Flutter cross-platform development |

| QA Agent | OWASP Top 10 security, performance, accessibility |

| Debug Agent | Bug diagnosis, root cause analysis, regression tests |

| Orchestrator | CLI-based parallel agent execution with Serena Memory |

| Commit | Conventional Commits with project-specific rules |

Prerequisites

• Google Antigravity (2026+)
• Bun (for CLI and dashboards)
• uv (for Serena setup)

Option 1: Interactive CLI (Recommended)

```bash

# Install bun if you don't have it:

# curl -fsSL https://bun.sh/install | bash

# Install uv if you don't have it:

# curl -LsSf https://astral.sh/uv/install.sh | sh

bunx oh-my-ag

```

Select your project type and skills will be installed to .agent/skills/.

| Preset | Skills |

|--------|--------|

| ✨ All | Everything |

| 🌐 Fullstack | frontend, backend, pm, qa, debug, commit |

| 🎨 Frontend | frontend, pm, qa, debug, commit |

| ⚙️ Backend | backend, pm, qa, debug, commit |

| 📱 Mobile | mobile, pm, qa, debug, commit |

Option 2: Global Installation (For Orchestrator)

To use the core tools globally or run the SubAgent Orchestrator:

```bash

bun install --global oh-my-ag

```

You'll also need at least one CLI tool:

| CLI | Install | Auth |

|-----|---------|------|

| Gemini | bun install --global @anthropic-ai/gemini-cli | gemini auth |

| Claude | bun install --global @anthropic-ai/claude-code | claude auth |

| Codex | bun install --global @openai/codex | codex auth |

| Qwen | bun install --global @qwen-code/qwen | qwen auth |

Option 3: Using vercel-labs/skills

```bash

bunx skills add first-fluke/oh-my-ag

```

Option 4: Clone & Open

```bash

git clone https://github.com/first-fluke/oh-my-ag

cd oh-my-ag

antigravity open .

```

Antigravity automatically detects skills in .agent/skills/.

Option 5: Integrate into Existing Project

Recommended (CLI):

Run the following command in your project root to automatically install/update skills and workflows:

```bash

bunx oh-my-ag

```

> Tip: Run bunx oh-my-ag doctor after installation to verify everything is set up correctly (including global workflows).

Manual Method:

If you have cloned this repo, you can simply copy the .agent directory:

```bash

# 1. Copy skills

cp -r oh-my-ag/.agent/skills /path/to/your-project/.agent/

# 2. Copy workflows

cp -r oh-my-ag/.agent/workflows /path/to/your-project/.agent/

# 3. Copy config (optional)

cp -r oh-my-ag/.agent/config /path/to/your-project/.agent/

```

2. Initial Setup (Optional)

```

/setup

→ Check CLI installations, MCP connections, configure language & CLI mapping

```

This creates .agent/config/user-preferences.yaml for your project.

3. Chat

Simple task (single agent auto-activates):

```

"Create a login form with Tailwind CSS and form validation"

→ frontend-agent activates

```

Complex project (workflow-guide coordinates):

```

"Build a TODO app with user authentication"

→ workflow-guide → PM Agent plans → agents spawned in Agent Manager

```

Explicit coordination (user-triggered workflow):

```

/coordinate

→ Step-by-step: PM planning → agent spawning → QA review

```

Commit changes (conventional commits):

```

/commit

→ Analyze changes, suggest commit type/scope, create commit with Co-Author

```

3. Monitor with Dashboards

```bash

bunx oh-my-ag dashboard # Terminal dashboard (bash)

bunx oh-my-ag dashboard:web # Web dashboard (Node.js)

# → http://localhost:9847

```

Progressive Disclosure

You don't manually select skills. Antigravity automatically:

1. Scans your chat request
2. Matches against skill descriptions in .agent/skills/
3. Loads the relevant skill only when needed
4. Saves tokens via lazy loading

Agent Manager UI

For complex projects, use Antigravity's Agent Manager (Mission Control):

1. PM Agent creates a plan
2. You spawn agents in the Agent Manager UI
3. Agents work in parallel with separate workspaces
4. Monitor progress via inbox notifications
5. QA Agent reviews the final output

SubAgent Orchestrator (CLI)

For programmatic parallel execution:

```bash

# Single agent

oh-my-ag agent:spawn backend "Implement auth API" session-01 ./backend

# Parallel agents via orchestrator skill

oh-my-ag agent:spawn backend "Implement auth API" session-01 ./backend &

oh-my-ag agent:spawn frontend "Create login form" session-01 ./frontend &

wait

```

Supports multiple CLI vendors: Gemini, Claude, Codex, Qwen

Multi-CLI Configuration

Configure different CLIs per agent type in .agent/config/user-preferences.yaml:

```yaml

# Response language

language: ko # ko, en, ja, zh, ...

# Default CLI (single tasks)

default_cli: gemini

# Per-agent CLI mapping (multi-CLI mode)

agent_cli_mapping:

frontend: gemini

backend: codex

mobile: gemini

pm: claude

qa: claude

debug: gemini

```

CLI Resolution Priority:

1. --vendor command line argument
2. agent_cli_mapping from user-preferences.yaml
3. default_cli from user-preferences.yaml
4. active_vendor from cli-config.yaml (legacy)
5. Hardcoded fallback: gemini

Run /setup to configure interactively.

Serena Memory Coordination

The Orchestrator writes structured state to .serena/memories/:

| File | Purpose |

|------|---------|

| orchestrator-session.md | Session ID, status, phase |

| task-board.md | Agent assignments and status table |

| progress-{agent}.md | Per-agent turn-by-turn progress |

| result-{agent}.md | Co