GameDesignOS skill
Seven skills and a local CLI for game design: from idea to pitch and experiments, with decisions approved by a human
- 437 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The CLI creates a workspace and writes decisions, assumptions, evidence and experiments into it, so it changes files on disk
- The skills need your material: gameplay recordings, notes and draft documents, including non-public projects
- Web research is used when the task requires it, so part of your data may end up in queries
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/gamedesignosDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The CLI creates a workspace and writes decisions, assumptions, evidence and experiments into it, so it changes files on disk
- The skills need your material: gameplay recordings, notes and draft documents, including non-public projects
- Web research is used when the task requires it, so part of your data may end up in queries
This is third-party code. Review the repository files before installing.
What it does
GameDesignOS is a set of seven skills and a local runtime for game design with an AI agent. The skills turn a game idea into a verifiable plan, turn a gameplay recording or trailer into a timestamped report, assemble a publisher pitch, plan weekly experiments on retention, pacing and feedback, translate chapters of game design books and maintain a source base. Results land in a project workspace: decisions, assumptions, evidence and experiments are linked, and skills exchange data through 19 JSON schemas. Decisions that change the project's commitments are recorded by a person: the system stops at a checkpoint called the Human Gate and waits for your decision, and it can roll back its own changes. The local `gamedesignos` CLI installs from a clone with `python -m pip install -e .`. The `python -m gamedesignos demo` command creates a training workspace with no model calls and no keys, and `python -m gamedesignos ask` picks a fitting skill and by default writes nothing to disk.
Who it is for. For game designers, producers and indie founders who work with Claude Code or Codex and want to keep a project's decisions and evidence in one place.
Good fit when
- You have a one-line game idea and need a player promise, a core loop and a prototype validation plan
- You have a gameplay recording or trailer and need a report with timestamps, evidence and issue priorities
- You need a publisher pitch with scope boundaries, budget assumptions and a decision request
- First-session retention is sagging and you need a week of A/B experiments with metrics and rollback rules
- A chapter of a game design book needs translating into Chinese with terminology that can be reviewed
Not a fit when
- You need game code or engine setup: the skills work on design, research and decisions
- You want a finished design document without checking assumptions: the skills analyze assumptions first and write afterwards
- You need the agent itself to decide the project's direction: in this system a person records commitments at the Human Gate checkpoint
- The project is private while the repository is public: keep personal material out of it, in your own workspace
Example request
Use game-concept-architect: I have a tactics game idea about a lighthouse keeper; break it into a player promise, a core loop and a prototype validation planLimitations
- Skills are installed by copying the folders you need into your agent's skills directory. There is no ready install command for Claude Code. Check that relative links in references, templates and examples still work.
- Skill instructions are written in Chinese, and the documentation comes in English and Chinese. Book translation targets Chinese as the output language.
- The CLI and runtime need Python and install from a clone of the repository. Public examples in the repository are synthetic or marked as needing a source review.
- Web research and writing to project files turn on only when the task needs them.
- The `python -m gamedesignos ask` command picks a skill but starts writing to a workspace only after you give it a destination folder.
How to disable. Delete the copied skill folders from the agent's skills directory. The installed CLI is removed with `python -m pip uninstall gamedesignos`. Workspaces you created sit in the folder you chose and go away with it.
FAQ
How do I install GameDesignOS skills in Claude Code?
Clone the repository and copy the skill folders you need, such as game-concept-architect, into the Claude Code skills directory. After copying, check that the name in SKILL.md matches the folder name. Notes for Claude Code are in adapters/claude-code.md.
What is the Human Gate in GameDesignOS?
It is a stopping point before a decision that changes the project's commitments. The agent proposes and structures the work, and a person records the decision. The training demo also stops before the decision accept command.
Do I need an API key to try GameDesignOS?
No. The `python -m gamedesignos demo` command makes no model calls and needs no key. It creates a Lighthouse training workspace in the system temp folder and runs up to the decision a person confirms.
Which skills does GameDesignOS include?
Seven: game-concept-architect, game-experience-analyzer, game-experience-density-optimizer, game-design-proposal-writer, paranoia-ai-system-evolver, game-design-book-translator and game-design-source-curator.
How do I create a persistent workspace for my own project?
Run `python -m gamedesignos start "Lighthouse Tactics" --destination ../lighthouse-designos` with your project's name. The runtime creates nine lifecycle sections: inbox, decisions, assumptions, evidence, experiments, design assets, workflows, learning and exports. Runtime state is kept separately in .gamedesignos/.
Discussion
No questions or reviews yet.
Related
SkillsA set of game-dev skills for AI agents with a router that picks the right ones for the engine and task
A Claude Code project template for game development: a studio hierarchy of agents, skills for every phase, hooks, rules and doc templates
GD Agentic Skills
gd-agentic-skills
Skills for Godot 4: the Orchestrator master skill and narrow skills by area. A strict warning: never install with the --all flag, only one at a time
A Claude Code marketplace of 19 plugins and 76 skills for frontend, backend, DevOps and security, with a skill chain command
skill-creator: skills from OpenAPI, GraphQL and MCP
@asnd/skill-creator
The /skill-creator command turns a link to an OpenAPI spec, GraphQL schema or MCP server into a ready skill with wrappers.
Autonomous game development for Godot, Bevy and Babylon.js: the agent writes code, generates assets, runs the engine and judges the result from the game
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| GameDesignOS (this entry) | no | 437 | |
| Claude Code Game Studios | no | 28,502 | a game studio template with an agent hierarchy and hooks, while GameDesignOS tracks a project's design decisions, evidence and experiments |
| awesome-gamedev-agent-skills | no | 5,447 | a game dev skill set with a router by engine and task, while GameDesignOS focuses on design and keeps decisions in the project workspace |
README in short
The page leads straight to a quick start with a demo and explains why the runtime exists: AI drafts faster than a team can decide, and scattered notes pile up. It also holds ready prompts for each skill and two public example reports with sources.
README badge
Original SKILL.md
--- name: game-concept-architect description: 将一句话游戏创意扩展为可验证的游戏设计蓝图：从 concept seed、玩家动词、design nucleus options、动作-目标对齐、assumption ledger、玩家承诺、核心循环、scope gate 到 validation plan。适用于真实项目、私有项目、客户项目、公开案例或 synthetic cases；仓库示例发布规则由 CONTRIBUTING.md 管理。 license: MIT compatibility: 需要读取用户提供的项目资料；联网调研与写入项目文件仅在任务明确需要时使用。 metadata: version: "1.3.0-candidate" short-description: 将游戏创意编译为可验证的设计蓝图 --- # Game Concept Architect 使用本 skill 时，把一句话游戏创意转化为可以判断、可以裁剪、可以原型验证的游戏设计蓝图。 不要把自己当成普通 GDD 生成器。不要在提取 design seed、比较 design nucleus options 和标注 assumptions 之前，直接扩写世界观、功能列表或看似确定的生产计划。 用户可在自己的环境中处理真实项目、私有项目或客户项目。准备公开仓库案例时，只能使用 synthetic、公开或明确授权的材料，并在发布前执行 Human Gate。 ## 强制顺序 在输出任何设计案之前，必须按下面顺序完成。用户要求简短时可以压缩，但不要跳过 gate。 1. `concept seed extraction` 2. `player verb inventory` 3. `design nucleus options` 4. `action-goal alignment` 5. `assumption ledger` 6. `player promise` 7. `core loop` 8. `scope gate` 9. `validation plan` 当输入涉及参考游戏、机制迁移、完整方案、玩法审查、受众动机、随机性、长期内容或主题表达时，启用 `game dissection lens`。只展开会改变关键设计判断的层级，不要为了显得完整而把所有模板都塞进输出。 ## 默认工作流 1. 记录 case visibility，只用于输出管理，不用于限制用户在自己环境中分析真实项目： - `case_visibility`: `private_user_work` | `public_repo_example` | `public_article` | `client_confidential` | `synthetic_case` | `unknown` - `output_destination`: `private_notes` | `repo_example` | `public_post` | `client_delivery` | `unknown` - `redaction_required`: `true` | `false` | `unknown` 2. 用一句话复述用户原始创意。 3. 提取 concept seed： - 题材母体 - 玩法母体 - 情绪承诺