unity-ai-workflow
Health Pass
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 10 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
AI-first Unity 6.2+ game development workflow — rules, agents, skills, and slash commands for Claude Code and Antigravity
Unity AI Workflow
A comprehensive, skills-first toolkit for AI-assisted Unity development, built for Claude Code. This repository provides the "brain" (skills, commands, docs, and templates) to turn Claude Code into an expert Unity team.
No Unity project inside — this repo is pure workflow infrastructure. You copy it into your game's workspace and Claude Code reads it as its operating manual.
Dev Modes
Choose how much the AI does vs. how much you do:
| Mode | Who Builds | Best For |
|---|---|---|
| Guided (default) | Collaborative. AI suggests; you confirm key decisions. | Most projects |
| Autonomous | AI builds after a short onboarding Q&A. | Rapid prototyping, jam games |
Set your mode in docs/ProjectConfig.yaml -> ai_mode. Enable Claude Code auto mode alongside autonomous for best results.
Core Philosophy: Game Feel is Not Optional Later
Every feature is built complete using /uw-cmd-implement-feature:
interrogation -> GDD/GFD docs -> TDD implementation -> game feel integration -> commit
The AI asks about VFX, SFX, camera reactions, and haptics before writing a single line of code. Polish is iterative, not a phase. A feature isn't "done" when it compiles — it's done when it plays well.
Recommended Workspace Structure
[!IMPORTANT]
Don't open the Unity project directly. Create a parent workspace folder containing both your Unity project and docs. This lets Claude Code read design documents, templates, AND Unity code simultaneously.
MyGame/ <- Open THIS folder in Claude Code
├── CLAUDE.md <- Auto-loaded instructions (minimal, ~80 lines)
├── .claude/
│ ├── commands/ <- Slash commands (/uw-cmd-implement-feature, /uw-cmd-brainstorm, etc.)
│ ├── skills/ <- 13 reusable knowledge modules
│ └── settings.json <- Hooks & permissions
├── docs/ <- Living documents (filled during development)
│ ├── ProjectConfig.yaml
│ ├── GDD.md, TDD.md, GFD.md
│ └── phases/ <- 6 phase guides
├── templates/ <- Pristine starters (for reference)
└── MyGameUnity/ <- Unity project (created via Unity Hub)
├── Assets/
├── Packages/
└── ProjectSettings/
Quick Start
- Create your workspace: Create a parent folder for your game (e.g.,
MyGame/). - Copy the toolkit: Copy everything from this repo into your workspace root.
- Create your Unity project: Inside the workspace via Unity Hub (e.g.,
MyGame/MyGameUnity/). - Open the workspace in your terminal or VS Code with Claude Code.
CLAUDE.mdis auto-loaded — rules and workflow suggestions are built in.- Run
/uw-cmd-setup-project: Claude Code walks you through dev mode selection and full project initialization.
How It Works
- Minimal CLAUDE.md: ~80 lines of non-discoverable Unity gotchas — only what the AI consistently gets wrong. No bloat. Based on research showing verbose context files reduce AI performance.
- Skills 2.0: 13 reusable knowledge modules in
.claude/skills/, using Anthropic's latest skill format with progressive disclosure (metadata -> body -> references). - Slash Commands: 9 workflows as Claude Code commands —
/uw-cmd-brainstorm,/uw-cmd-plan,/uw-cmd-create,/uw-cmd-implement-feature,/uw-cmd-test,/uw-cmd-debug,/uw-cmd-polish,/uw-cmd-review,/uw-cmd-setup-project. - Verification System: Every AI recommendation is marked
[VERIFIED],[SYNTHESIZED], or[UNVERIFIED]— no silent hallucinations. - Integrated Feature Loop:
/uw-cmd-implement-featurecombines code + game feel in one pass, with deep interrogation before any coding starts. - Session Handoff:
SessionStatetemplate ensures context survives across conversations. - Hooks: Claude Code hooks in
.claude/settings.jsonfor automated workflows.
Skills
| Skill | Purpose |
|---|---|
uw-game-feel-integrator |
Inject juice (VFX, SFX, camera, tweens, haptics) using Rule of Three |
uw-scriptable-object-arch |
SO-first data architecture (default pattern) |
uw-dependency-injection |
DI-first using Reflex (opt-in) |
uw-state-machine |
Interface-based FSMs for game flow and characters |
uw-ui-toolkit-binder |
Unity 6+ Runtime Data Binding (MVVM) |
uw-network-setup |
Package-agnostic networking boilerplate |
uw-unity-project-setup |
Full project initialization wizard |
uw-unity-feature-scaffold |
Feature folder + asmdef + test assembly |
uw-unity-test-runner |
NUnit test generation from Gherkin scenarios |
uw-unity-debugging |
GameDebug wrapper with conditional compilation |
uw-unity-editor-tools |
Custom Inspectors, Gizmos, EditorWindows |
uw-code-review |
Unity-specific pre-commit quality checklist |
uw-skill-creator |
Official Anthropic skill-creator 2.0 — create, test, and optimize new skills |
Project Phases
Follow the phase guides in docs/phases/ for a standardized development journey:
- 00: Ideation — From rough idea to GDD + GFD.
- 01: Pre-Production — Technical choices, stack definition, naming conventions.
- 02: Technical Design — Architecture, Assembly Definitions, patterns.
- 03: Project Setup — Automated folder scaffolding and package installation.
- 04: Production — Feature-by-feature loop (interrogate -> implement -> feel -> commit).
- 05: Polish — Final tuning pass, visual refinement, performance profiling.
Configuration
The AI reads docs/ProjectConfig.yaml to stay in sync with your versions, packages, and architectural choices. Key settings:
ai_mode: "guided" # guided | autonomous
prototype_mode: false # Relaxes architecture rules for rapid iteration
Roadmap
See ROADMAP.md for the full version history and what's planned next.
Resources & Inspiration
This workflow was built with the help of these resources. Ratings reflect how much each influenced the final result.
Key Influences
| Resource | Rating | Influence |
|---|---|---|
| anthropics/skills | ***** | Official skill format — canonical reference, skill-creator 2.0 |
| Common-ka/ai-agent-unity-rules | **** | Unity-specific AI rules — shaped rules structure |
| obra/superpowers | **** | Composable skill format, agent routing philosophy |
| gsd-build/get-shit-done | **** | Pragmatic workflow philosophy, session state |
| CoplayDev/unity-mcp | **** | Unity MCP integration |
| stillwwater/UnityStyleGuide | **** | Naming conventions baseline |
| PracticAPI — The Perfect Unity Project | ***** | Reference architecture for production Unity 6 |
| Loic Jacob — Game Feel Theory | ***** | ADSR envelope, Three Pillars, Sensation Vocabulary |
| TCREI Prompt Engineering Framework | ***** | TCREI methodology baked into commands |
Research That Shaped v2.0
| Resource | Key Insight |
|---|---|
| Addy Osmani — AGENTS.md | Verbose context files reduce AI success rates. Only non-discoverable info belongs. |
| Zazen Codes — Stop using AGENTS.md | Keep files short, eliminate duplication, agent-specific instructions only |
| Claude Blog — Auto Mode | Native auto mode replaces custom autonomous dev modes |
| Claude Blog — Code Review | Multi-agent parallel review, GitHub App integration |
| Claude Blog — Hooks | 8 hook types for workflow automation |
| Claude Blog — Skills | Evaluation-driven skill development with quantitative benchmarks |
Created by devdavv. Targeted for Unity 6.2+ and modern AI workflows.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found