Agentic Creator OS

The Framework for AI-Empowered Creative Transformation

"Transform from overwhelmed creator to confident AI-empowered artist"

┌─────────────────────────────────────────────────────────────────────────┐
│                         AGENTIC CREATOR OS                              │
│                                                                          │
│   An open framework for building personal AI creative systems that      │
│   work WITH existing agent tools (Claude, OpenCode, etc.), not          │
│   instead of them.                                                      │
│                                                                          │
│   ┌─────────────┐  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐   │
│   │  FRAMEWORK  │  │    SKILLS   │  │   PROTOCOL  │  │   TEMPLATES │   │
│   │  (This Doc) │  │   Library   │  │  & Handoffs │  │  & Configs  │   │
│   └─────────────┘  └─────────────┘  └─────────────┘  └─────────────┘   │
│                                                                          │
│   Integrates with: Claude CLI • OpenCode • Codex • Gemini               │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

---

Table of Contents

1. Philosophy
2. Architecture
3. Core Components
4. Agent Team Design
5. Skills System
6. Multi-Agent Protocol
7. Brand Voice Framework
8. Configuration Guide
9. Creating Your Instance
10. Examples

---

Philosophy

Creator-First AI

┌──────────────────────────────────────────────────────────────────────┐
│  TRADITIONAL APPROACH              │  CREATOR-FIRST APPROACH         │
├────────────────────────────────────┼──────────────────────────────────┤
│  AI does the work                  │  AI amplifies human creativity  │
│  Features over feelings            │  Soul + Systems                 │
│  Corporate jargon                  │  Studio energy                  │
│  One-size-fits-all                │  Personalized workflows         │
│  "Leverage the platform"           │  "What do you want to create?"  │
└────────────────────────────────────┴──────────────────────────────────┘

Core Principles

1. Empower, Don't Replace - Every AI interaction should leave the creator more capable, not dependent
2. Soul + Systems - Bridge spirituality and technology; technical excellence with human warmth
3. Action-Oriented - Every deliverable ends with a clear next step
4. Studio Energy - Communicate with warmth and intimacy, not boardroom presentations

The Transformation Promise

"From overwhelmed creator to confident AI-empowered artist"

This isn't about faster coding or more content. It's about the fundamental shift from:

• Consumer → Creator
• Dependent → Capable
• Overwhelmed → Empowered
• Alone → Supported

---

Architecture

Layered Design

┌─────────────────────────────────────────────────────────────────────────┐
│                         LAYER 1: FRAMEWORK                              │
│              The open, shareable part (Agentic Creator OS)              │
├─────────────────────────────────────────────────────────────────────────┤
│  • Core philosophy & principles                                         │
│  • Agent team templates                                                 │
│  • Skills architecture                                                  │
│  • Protocol definitions                                                 │
│  • Configuration schemas                                                │
└─────────────────────────────────────────────────────────────────────────┘
                                    ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                        LAYER 2: INSTANCE                                │
│              Your personal customization (e.g., FrankX)                 │
├─────────────────────────────────────────────────────────────────────────┤
│  • Your brand voice & style                                             │
│  • Your specific skills                                                 │
│  • Your agent team configuration                                        │
│  • Your workflows & projects                                            │
│  • Your magic words & patterns                                          │
└─────────────────────────────────────────────────────────────────────────┘
                                    ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                        LAYER 3: EXECUTION                               │
│              The agent tools that run your instance                     │
├─────────────────────────────────────────────────────────────────────────┤
│  • Claude CLI (Claude Code)                                             │
│  • OpenCode CLI (with oh-my-opencode)                                   │
│  • Google Gemini CLI                                                    │
│  • OpenAI Codex CLI                                                     │
└─────────────────────────────────────────────────────────────────────────┘

Agent System Compatibility

System	Best For	Integration Method
Claude CLI	Complex reasoning, writing, brand voice	Agent specs (*.md in ~/.claude/agents/)
OpenCode	Rapid prototyping, parallel execution	oh-my-opencode config (.opencode/oh-my-opencode.json)
Codex	Enterprise architecture, OCI integration	Agent specs
Gemini	Web implementation, QA, accessibility	Agent specs

---

Core Components

1. The Orchestrator Agent

Every instance needs a primary orchestrator. This agent:

• Receives initial requests
• Coordinates subagents
• Ensures brand consistency
• Manages handoffs

Default Name: Sisyphus (oh-my-opencode default) Customizable: Yes, via prompt_append

2. Subagent Team

┌─────────────────────────────────────────────────────────────────────────┐
│                        STANDARD SUBAGENT TEAM                           │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐   │
│  │  LIBRARIAN  │  │   EXPLORE   │  │   ORACLE    │  │  FRONTEND   │   │
│  │  Researcher │  │   Scanner   │  │   Advisor   │  │   Designer  │   │
│  └─────────────┘  └─────────────┘  └─────────────┘  └─────────────┘   │
│                                                                          │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐                     │
│  │  DOCUMENT   │  │  MULTIMODAL │  │   BUILDER   │                     │
│  │   Writer    │  │   Analyzer  │  │   Coder     │                     │
│  └─────────────┘  └─────────────┘  └─────────────┘                     │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

3. Skills Library

Modular knowledge units that agents can invoke. See Skills System.

4. Brand Voice Layer

Consistent tone, vocabulary, and messaging across all outputs. See Brand Voice Framework.

---

Agent Team Design

Template: Core Agent Roles

## ORCHESTRATOR (Primary Agent)

**Role**: System coordinator and primary interface
**Personality**: 
- Welcoming and clarifying
- Ensures all subagents align with system goals
- Maintains coherence across complex tasks

**When to Activate**: Every request starts here

---

## LIBRARIAN (Subagent)

**Role**: Research and information retrieval
**Personality**: 
- Thorough and precise
- Finds the elegant solution, not just any solution
- Brings wisdom, not just information

**When to Activate**: Need to research, find docs, look up APIs

---

## EXPLORE (Subagent)

**Role**: Codebase navigation and pattern discovery  
**Personality**:
- Pattern-focused
- Identifies opportunities and anti-patterns
- Maps the territory

**When to Activate**: Need to understand codebase, find patterns, locate files

---

## ORACLE (Subagent)

**Role**: Strategic advice and architecture decisions
**Personality**:
- Deep thinking
- Long-term perspective
- Bold recommendations with clear reasoning

**When to Activate**: Complex decisions, architecture reviews, strategic planning

---

## FRONTEND-UI-UX-ENGINEER (Subagent)

**Role**: Interface design and implementation
**Personality**:
- Design-conscious
- Accessibility-focused
- User empathy

**When to Activate**: UI components, design systems, user interfaces

---

## DOCUMENT-WRITER (Subagent)

**Role**: Documentation and content creation
**Personality**:
- Clear and organized
- Reader-empathetic
- Long-term thinking

**When to Activate**: Documentation, guides, content writing

---

## MULTIMODAL-LOOKER (Subagent)

**Role**: Visual media analysis
**Personality**:
- Interpretive
- Context-aware
- Insightful

**When to Activate**: Analyzing images, diagrams, screenshots

Customizing Your Team

You can customize any agent by adding:

1. Custom Persona (via prompt_append in oh-my-opencode.json)
2. Custom Skills (add to skills library)
3. Custom Tools (via MCP servers)
4. Custom Models (per-agent model selection)

---

Skills System

Skills Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                           SKILLS DIRECTORY                              │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  .claude-skills/                                                        │
│  ├── technical/           # Technical/domain skills                      │
│  │   ├── mcp-architecture/                                             │
│  │   ├── nextjs-react-expert/                                           │
│  │   └── ...                                                            │
│  ├── creative/            # Creative/content skills                      │
│  │   ├── brand-voice/                                                │
│  │   ├── content-creation/                                             │
│  │   └── ...                                                            │
│  ├── personal/            # Personal development skills                  │
│  │   ├── philosophy/                                                   │
│  │   ├── discipline/                                                    │
│  │   └── ...                                                            │
│  ├── business/            # Business/product skills                      │
│  ├── projects/            # Project-specific skills                      │
│  └── soulbook/            # Life design system (25+ skills)              │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

Skill File Format

---
name: Skill Name
description: Brief description of what this skill does
version: 1.0.0
category: technical|creative|personal|business|projects
---

# Skill Name

## Purpose
What this skill helps you accomplish.

## Core Knowledge
Key concepts, patterns, and best practices.

## When to Use
Situations where this skill applies.

## Examples
Practical examples and templates.

## Integration
How to invoke this skill in your workflow.

Using Skills

System	How to Invoke
Claude Code	Auto-activates via hooks, or use /skill [name]
OpenCode	Read the SKILL.md file directly
Codex	Read the SKILL.md file directly
Gemini	Read the SKILL.md file directly

---

Multi-Agent Protocol

Handoff Template

## Handoff: [Feature/Task Name]
**From**: [Agent Name]
**To**: [Agent Name]
**Status**: [Ready / Needs Review / Blocked]

### What's Done
- [Completed work items]

### What's Needed
- [What the next agent should do]

### Files Changed
- [List of modified files]

### Context
- [Important decisions, constraints, notes]

### Quality Notes
- [Any quality concerns or areas needing attention]

Priority Levels

Level	Description	Response Time
P0	Production down	Immediate
P1	Major feature blocked	Same session
P2	Normal development	Next session
P3	Nice to have	Backlog

Coordination Patterns

Pattern 1: Sequential Handoff

User → Orchestrator → Specialist 1 → Orchestrator → Specialist 2 → User

Pattern 2: Parallel Execution

User → Orchestrator → [Specialist 1, Specialist 2, Specialist 3] → Synthesize → User

Pattern 3: Consultation

User → Orchestrator → Specialist (for expertise) → Back to Orchestrator → User

---

Brand Voice Framework

Voice Attributes

Attribute	Description	Example
Cinematic	Paint vivid pictures	"The studio lights snap on at 2 AM"
Intimate	Direct to individual	"You know that moment when..."
Studio-Rooted	Creative practice grounding	"Like arranging a track..."
Technically Warm	Complex but accessible	"Like a producer balancing stems"

Words to Use

Create, craft, build, compose, architect, orchestrate
Flow, rhythm, resonance, frequency, vibe
Transform, empower, amplify, elevate
Session, ritual, practice, journey
Authentic, soul-aligned, conscious

Words to Avoid

Disrupt, revolutionary, game-changing (startup jargon)
Synergy, leverage, utilize (corporate speak)
Cheap, easy, simple (undermines craft)
Automate everything (we augment, not replace)
Just, merely, simply (condescending)

Tone by Context

Context	Tone	Example Lead
Blog/Content	Story first, system second	"Picture this: 2 AM in the studio..."
Technical	Warm precision	"Think of it like mixing stems..."
Social	Punchy, inspiring	"Most creators overthink."
Email	Conversational, actionable	"Quick note from the studio..."

---

Configuration Guide

OpenCode (oh-my-opencode) Configuration

Location: .opencode/oh-my-opencode.json

{
  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json",
  "agents": {
    "Sisyphus": {
      "model": "anthropic/claude-opus-4-5",
      "prompt_append": "\n\n## IDENTITY OVERRIDE\nYou are [YOUR NAME] - [Your mission].\n\n[Your brand voice guidelines...]\n",
      "permission": {
        "edit": "allow",
        "bash": "ask"
      }
    },
    "librarian": {
      "model": "opencode/glm-4.7-free",
      "prompt_append": "\n\n[Your customization...]"
    },
    "explore": {
      "model": "opencode/glm-4.7-free",
      "prompt_append": "\n\n[Your customization...]"
    },
    "oracle": {
      "model": "anthropic/claude-opus-4-5",
      "prompt_append": "\n\n[Your customization...]"
    }
  }
}

Claude Code Agent Specifications

Location: ~/.claude/agents/[agent-name].md

# [Agent Name]
*[Role/Tagline]*

## Mission
[Brief mission statement]

## Core Identity
- **Role**: [Primary function]
- **Personality**: [Key traits]
- **Thinking Style**: [How they approach problems]

## Primary Responsibilities
- [Responsibility 1]
- [Responsibility 2]
- [Responsibility 3]

## Skills Access
```bash
/skill [skill-1]
/skill [skill-2]
/skill [skill-3]

Coordination

Agent	Role	Handoff When
[Other Agent]	[Their Role]	[Condition]

---

Work with [your brand] voice and principles.

---

## Creating Your Instance

### Step 1: Define Your Identity

**Answer these questions:**

1. What's your name? (e.g., FrankX, YourName)
2. What's your mission?
3. What's your core audience?
4. What's your unique voice?

**Template:**
```markdown
## MY INSTANCE IDENTITY

**Name**: [Your Name]
**Mission**: [What you help creators achieve]
**Audience**: [Who you serve]
**Voice**: [3-5 words describing your tone]

Step 2: Configure Your Agent Team

1. Choose your orchestrator agent (Sisyphus via OpenCode)
2. Customize each subagent with prompt_append
3. Set model preferences per agent
4. Define permissions

Step 3: Build Your Skills Library

1. Copy the .claude-skills/ structure
2. Add your domain-specific skills
3. Customize existing skills for your context
4. Document in SKILLS.md

Step 4: Document Your Workflows

## MY WORKFLOWS

### Content Creation
1. [Step 1]
2. [Step 2]
3. [Step 3]

### Book Writing
1. [Phase 1]
2. [Phase 2]
3. [Phase 3]

### Website Development
1. [Phase 1]
2. [Phase 2]
3. [Phase 3]

Step 5: Test and Iterate

1. Run your first complex task
2. Note where coordination breaks
3. Adjust configurations
4. Repeat

---

Examples

Example 1: FrankX (Personal Instance)

Location: /mnt/c/Users/Frank/FrankX/

FrankX is a personal instance for a musician-technologist helping creators.

Key Customizations:

• Studio energy voice (musician background)
• 52 skills including Soulbook life design
• Multi-agent team: Claude (Story), Codex (Architecture), Gemini (QA), OpenCode (Build)
• Magic words: ultrawork, ultracode

Files:

• CLAUDE.md - Claude agent spec
• OPENCODE.md - OpenCode agent spec
• GEMINI.md - Gemini agent spec
• AGENTS.md - Multi-agent coordination
• .opencode/oh-my-opencode.json - OpenCode config
• .claude-skills/ - 52 skills

Example 2: Arcanea (Fantasy World Instance)

Location: /mnt/c/Users/Frank/Arcanea/

Arcanea is an instance for a fantasy world-building system.

Key Customizations:

• Luminor character system (7 aspects)
• Future-era perspective (100 years ahead)
• Magic words for world-building

Example 3: Your Instance (Template)

your-project/
├── YOUR-NAME.md              # Your agent spec
├── SKILLS.md                 # Your skills
├── WORKFLOWS.md              # Your workflows
├── .opencode/
│   └── oh-my-opencode.json   # OpenCode config
└── .claude-skills/
    ├── technical/
    ├── creative/
    ├── personal/
    └── [your-categories]/

---

Magic Words

Magic words trigger parallel agent execution or special behaviors:

Word	Effect
ultrawork / ulw	Fire ALL available agents in parallel
ultracode / ulc	Fire coding agents (architect + coder + reviewer)
ultrawrite / ulwr	Fire writing agents in parallel
ultrabook / ulb	Complete book pipeline

---

Directory Structure

agentic-creator-os/
├── README.md                  # This file
├── CREATOR-OS.md              # Framework documentation
├── TEMPLATES/
│   ├── agent-spec.md          # Agent specification template
│   ├── skill-template.md      # Skill file template
│   ├── handoff-template.md    # Handoff template
│   └── workflow-template.md   # Workflow documentation template
├── CONFIGS/
│   ├── oh-my-opencode/
│   │   └── basic.json         # Basic oh-my-opencode config
│   └── claude-agents/
│       └── orchestrator.md    # Orchestrator agent spec
├── SKILLS/
│   ├── technical/
│   ├── creative/
│   ├── personal/
│   └── README.md              # Skills documentation
└── EXAMPLES/
    ├── frankx/                # FrankX as reference implementation
    └── [your-instance]/       # Your custom instance

---

Contributing

Agentic Creator OS is designed to be extended:

1. Add templates in TEMPLATES/
2. Share skills in SKILLS/
3. Document examples in EXAMPLES/
4. Improve the framework via pull requests

---

License

Open source under MIT License.

---

Acknowledgments

• Oh-My-OpenCode - The configuration system this framework builds upon
• Claude Code - Primary agent execution environment
• All AI Agent Frameworks - For inspiring the multi-agent future

---

The framework for AI-empowered creative transformation.

Agentic Creator OS - Create with soul. Build with systems.