Why This Exists

Most Claude Code setups are a single CLAUDE.md file with some instructions.
This studio gives Claude Code the structure of a real development team —
37 specialized agents, a guided onboarding sequence that takes any idea
from one sentence to a configured project, and a sprint system that
keeps builds disciplined from start to launch.

You bring the idea. The studio handles the rest.

Start Here

# Clone
git clone https://github.com/marko-builds/Claude-Code-Dev-Studio my-project
cd my-project && rm -rf .git && git init

# Open Claude Code
claude

# That's it. The studio takes it from here.
/start

You only need an idea. /start interviews you, challenges your concept, detects or builds the right track, produces all project documents, plans your sprints, and hands you a configured project.

Works for non-developers and experienced engineers alike.

What's In Here

Studio Brain (new)

The /start command is now a full orchestration sequence. From raw idea to configured project in one conversation:

1. Identifies whether you're a developer or idea-first founder — adapts accordingly
2. Interviews you about your concept (5-8 questions)
3. Challenges it honestly before writing anything (Go / Refine / Pivot)
4. Detects the right track — or generates a custom one if needed
5. Produces all documents with your approval at each step
6. Plans your full sprint roadmap
7. Hands off with "Sprint 0 starts now"

Design Phase (new in v3)

Take any idea from one sentence to a dependency-ordered sprint plan
before writing a single line of code.

/design → Guided ADD authoring — every system, every decision mapped
/map-systems → Dependency graph — optimal build order, no mid-project blockers
/design-index → Living decision record — every architectural choice with rationale

Full flow: /start → /design → /map-systems → /sprint-plan → /sprint-start 0

Three-Tier Agent Hierarchy (37 agents)

Tier 0: Studio Brain
  studio-brain, track-architect

Tier 1: Directors
  product-director, architecture-director, tech-lead

Tier 2: Department Leads
  frontend-lead, mobile-lead, backend-lead, devops-lead,
  design-lead, qa-lead, security-lead, launch-lead, content-lead, game-director

Tier 3: Specialists — Software Track
  frontend-developer, mobile-developer, backend-developer,
  database-architect, auth-specialist, payments-specialist,
  devops-engineer, code-reviewer, accessibility-specialist,
  performance-engineer, documentation-writer

Tier 3: Specialists — Game Dev Track
  game-designer, level-designer, narrative-director,
  unity-developer, godot-developer, unreal-developer,
  gameplay-programmer, ai-programmer, tools-programmer

Tier 3: Specialists — Content & Research
  content-writer, research-analyst

24 Commands

Studio:     /start /concept-review /new-track /dashboard /onboard
Sprints:    /sprint-plan /sprint-start /sprint-end
Building:   /spec /new-project /review /test /refactor /debug
Quality:    /audit /deploy /launch
Knowledge:  /wiki-update /wiki-query /reflect /research /docs /status

Tracks

20 Skills (auto-load by file context)

tdd-workflow, api-design, saas-patterns, mobile-patterns, auth-patterns, security-review, database-patterns, performance, sprint-methodology, concept-validation, launch-strategy, game-design-patterns, unity-patterns, godot-patterns, unreal-patterns, content-production, research-methodology, track-generation, systematic-debugging, two-stage-review

Karpathy Wiki (project + personal)

Every project gets a wiki/ directory — persistent, compounding knowledge maintained by /wiki-update:

• Architecture decisions with rationale
• Session-by-session log
• Gotchas and discoveries
• Open questions

Personal cross-project learning at ~/.claude/wiki/:

• Velocity data across all projects
• Recurring patterns and their fixes
• Studio improvement suggestions

6 Rescue Prompts

For difficult development moments: prompts/stuck-on-bug.md, scope-creep.md, handover.md, second-opinion.md, fresh-eyes.md, reset-session.md

Repository Layout

.
├── CLAUDE.md                     ← Master config
├── README.md                     ← This file
├── .claude/
│   ├── settings.json
│   ├── agents/     (37 agents)
│   ├── commands/   (24 commands)
│   ├── skills/     (20 skills)
│   ├── hooks/scripts/ (4 hooks)
│   └── rules/      (5 rule sets)
├── docs/
│   ├── COLLABORATIVE-DESIGN-PRINCIPLE.md
│   ├── SPRINT-METHODOLOGY.md
│   ├── TRACK-SYSTEM.md
│   ├── GAME-DEV-GUIDE.md
│   ├── STUDIO-BRAIN-GUIDE.md
│   ├── LEARNING-LOOP.md
│   ├── quick-start.md
│   ├── agent-roster.md
│   └── coordination-map.md
├── wiki/                         ← Project wiki scaffold
│   ├── dashboard.md
│   ├── index.md
│   ├── log.md
│   ├── open-questions.md
│   ├── architecture.md
│   ├── decisions/
│   └── discoveries/
├── prompts/                      ← Rescue prompts
│   ├── stuck-on-bug.md
│   ├── scope-creep.md
│   ├── handover.md
│   ├── second-opinion.md
│   ├── fresh-eyes.md
│   └── reset-session.md
└── templates/
    ├── web-saas/CLAUDE.md
    ├── mobile-app/CLAUDE.md
    ├── api-backend/CLAUDE.md
    ├── fullstack/CLAUDE.md
    ├── game/CLAUDE.md
    ├── content/CLAUDE.md
    └── research/CLAUDE.md

Philosophy

Collaborative, not autonomous. Every significant action: Question → Options → Decision → Draft → Approval → Execute.

Idea-first. You only need an idea. The studio handles classification, tech stack, documents, and sprint planning.

Prescriptive. The studio tells you what to build next, in what order, and why — not just gives you tools.

Compounding. Sessions build on each other. Projects build on each other. The studio gets smarter about you over time.

Credits

• Structural inspiration: Claude-Code-Game-Studios by Donchitos
• Wiki pattern: Andrej Karpathy's LLM-Wiki
• Hook + rules patterns: everything-claude-code by affaan-m
• Best practices: claude-code-best-practice by shanraisshan
• Community: awesome-claude-code by hesreallyhim
• Templates: claude-code-templates by davila7
• Workflow methodology: superpowers
  by Jesse Vincent — systematic debugging and two-stage review patterns

MIT License