DY-2026/GameDesignOS

What it solves

GameDesignOS addresses the problem of fragmented AI-generated game design output. Instead of scattered chat logs and one-off prompts, it provides a structured operating layer that turns AI suggestions into durable, reviewable project assets like decisions, assumptions, and evidence-linked experiments.

How it works

The system operates as a local-first CLI and workspace manager that integrates with AI agents. It uses a "Skill Kernel" of seven specialist workflows (e.g., concept architecture, experience analysis) and a "Contract Layer" of 19 schemas to ensure that outputs from different AI skills are interoperable and structured.

Workflows follow a strict path: Idea $\rightarrow$ Evidence $\rightarrow$ Experiment $\rightarrow$ Decision $\rightarrow$ Learning. To prevent AI from making critical project choices autonomously, the system implements "Human Gates," requiring a person to review and accept decisions before they are recorded.

Who it’s for

Game designers and development teams who use AI to accelerate their design process but need to maintain human authority, project memory, and evidence-based decision-making.

Highlights

• Evidence-First Approach: Every judgment points back to specific sources, such as timestamps in gameplay recordings or screenshots.
• Human-Gated Decisions: AI can propose and structure work, but commitment-changing decisions must be human-approved.
• Specialist Skill Kernel: Includes dedicated tools for concept architecture, experience density optimization, and proposal writing.
• Local-First Runtime: A deterministic CLI for managing workspaces, validating project health, and exporting decision graphs without requiring constant model calls.
• Agent Portable: Compatible with various AI agents (e.g., Claude Code, Codex) that support Markdown-based skills.