Claude Code Game Studio
A structured Claude Code team for planning, building, testing, and releasing games.
Per-dimension scores and reasoning
Evidence shows: README describes permission rules (settings.json auto-allows safe operations, blocks dangerous ones) and user confirmation process (collaboration protocol requires user decisions). However, data flow transparency is insufficient: no details on what data is collected, stored, or transmitted. Sensitive data handling only mentions blocking .env file reads, but not other sensitive data. Dependency security does not mention dependency auditing or vulnerability scanning. External effects: hooks execute shell commands, but README claims hooks exit immediately when irrelevant, and SECURITY.md provides security guidelines. Rollback: no explicit rollback mechanism, only upgrade guide. Source attribution: no clear attribution of third-party code. Therefore, some criteria scored lower.
Evidence shows: README and SECURITY.md are consistent in describing hook behavior, but no actual test results are provided. Dependency availability: no specific dependency versions listed, only recommends jq and Python 3, but no version requirements. Failure messages: hooks claim to degrade gracefully, but no specific error message examples. Therefore, self-consistency scored moderate, dependency availability and failure messages scored lower.
Evidence shows: README clearly identifies target audience (game developers) and multiple use scenarios (from brainstorming to release). Capability boundaries: describes agent responsibilities and domain boundaries, and collaboration protocol. Trigger precision: hooks and skills trigger conditions are clearly described. Environment fit: claims support for Windows, macOS, Linux, but primary testing on Windows 10, and cross-platform testing ongoing. Therefore, all criteria scored 2.
Evidence shows: README provides detailed project structure, installation steps, upgrade guide, and customization instructions. Naming stability: agent and skill names are consistent. Examples and FAQ: README provides usage examples, but no FAQ. Known limitations: README mentions cross-platform testing gaps and notification limitations. License: MIT license clear. Versioning and changelog: no CHANGELOG, but UPGRADING.md exists. Maintenance responsibility: README mentions maintenance and contribution, but no explicit maintainer. Therefore, some criteria scored lower.
Evidence shows: Output usability: README describes skill and agent outputs, but no actual output examples. Marginal value: provides 49 agents and 73 skills, potentially increasing productivity, but not quantified. Cost-benefit: no performance or resource consumption data. Therefore, all scored 2.
Evidence shows: Claim traceability: README claims (e.g., agent counts) are consistent with file structure, but no verification method. Cross-source corroboration: no external verification. Fact-inference separation: README distinguishes facts (e.g., agent counts) from design philosophy, but not explicitly labeled. Therefore, scores are lower.
- No actual test results provided; all claims based on static review, treat with caution.
- Data flow and sensitive data handling details are insufficient; recommend reviewing actual hook and skill code.
- Dependency versions not specified, potential compatibility issues.
- Cross-platform support not fully verified; primary testing on Windows 10.
What does this agent do, and when should you use it?
Claude Code Game Studios is a project template for Claude Code with 49 specialized agents, 73 slash-command skills, 12 hooks, 11 path-scoped rules, and 41 document templates. Its three-tier studio structure covers design, programming, art, audio, narrative, QA, and production, with separate specialist sets for Godot 4, Unity, and Unreal Engine 5. Users work through commands such as /start, /brainstorm, /dev-story, /qa-plan, and /release-checklist to produce design, architecture, planning, testing, and release artifacts. It is explicitly collaborative rather than autonomous: agents ask questions, present options and drafts, and require the user to decide and approve writing. The deployment boundary is a local project run in Claude Code, where repository configuration, hooks, and rules guide file operations and session workflow.
After cloning, the repository configures Claude Code through CLAUDE.md and the .claude/ agents, skills, hooks, rules, and docs directories. A user can run /start to select a project stage, or invoke skills including /setup-engine godot 4.6, /brainstorm, /create-epics, /create-stories, and /dev-story across the documented seven-phase pipeline. Agents use vertical delegation, peer consultation, escalation to creative-director or technical-director for conflicts, and producer-led coordination for cross-department changes. Hooks run at session open, compaction, close, agent start and stop, and relevant Bash or Write/Edit actions; validate-commit.sh, validate-assets.sh, and validate-skill-change.sh can check hardcoded values, JSON, asset naming, and skill changes. Path-scoped rules apply requirements to locations including src/gameplay/, src/core/, src/ai/, src/networking/, src/ui/, design/gdd/, tests/, and prototypes/.
- A solo developer with only a rough game idea who wants to start a guided design process in Claude Code with /start and /brainstorm.
- A small team using Godot 4, Unity, or Unreal Engine 5 that wants engine-specific specialist roles alongside general studio roles.
- A developer inheriting an existing prototype or codebase who wants to run /project-stage-detect and use /reverse-document to document it.
- A team planning iterative feature delivery with /create-epics, /create-stories, /sprint-plan, /dev-story, and /story-done.
- A project lead approaching test or release work who needs /qa-plan, /regression-suite, /release-checklist, /launch-checklist, and /day-one-patch.
- A creator who wants coordinated multi-role prompts for a feature, while retaining approval authority through skills such as /team-combat, /team-ui, and /team-qa.
What are this agent's strengths and limitations?
- Defines a 49-agent, three-tier studio hierarchy with explicit escalation and cross-department change-propagation paths.
- Supplies 73 slash-command workflows spanning ideation, design, architecture, implementation, QA, production, and release instead of a single general prompt.
- Includes 12 automated hooks and 11 path-scoped rules for commit, push, asset, session-lifecycle, and code-area checks.
- Offers distinct specialist agent sets for Godot 4, Unity, and Unreal Engine 5, with the option to use none of them.
- Its documented runtime is Claude Code; the repository provides no native support path for ChatGPT, Codex, the OpenAI API, or the Claude API.
- The template is broad, with many agents, skills, hooks, rules, and document templates that adopters may need to prune or tailor.
- Windows 10 is the primary development and test platform; macOS and Linux support is described with caveats, ongoing cross-platform testing, and a notification hook that is inactive outside Windows.
- Without the recommended jq or Python 3 tools, hooks fail gracefully but their corresponding validation is unavailable.
How do you install or deploy this agent?
Prerequisites are Git and Claude Code; jq is recommended for hook validation and Python 3 for JSON validation. Run: git clone https://github.com/Donchitos/Claude-Code-Game-Studios.git my-game
cd my-gameclaude
The documented Claude Code installation command is: npm install -g @anthropic-ai/claude-code. The repository does not document any additional credential requirement.
How do you use this agent?
Start claude in the project directory, then run /start and answer the project-stage prompt. If you already know your next step, use /brainstorm, /project-stage-detect, or /setup-engine godot 4.6. Continue with the design, architecture, story, QA, and release skills that fit the work. Review the agents’ options and drafts, then make the decisions and approvals; the documented protocol says nothing is written without user sign-off.