Agent-Skills.md

Agent Skills: Game Architect Skill

A documentation-focused skill for game architecture design. Produces technical selection, design, and planning documents through a structured pipeline. Use this skill to generate requirement analysis, technical design, and implementation planning documents for new game projects or major feature development.

UncategorizedID: Yuki001/game-dev-skills/game-architect

Repository

Yuki001/game-dev-skills
Yuki001
1

Install this agent skill to your local

pnpm dlx add-skill https://github.com/Yuki001/game-dev-skills/game-architect

Skill Files

Browse the full folder contents for game-architect.

Download Skill

Loading file tree…

skills/game-architect/SKILL.md

Skill Metadata

Name
game-architect
Description
A documentation-focused skill for game architecture design. Produces technical selection, design, and planning documents through a structured pipeline. Use this skill to generate requirement analysis, technical design, and implementation planning documents for new game projects or major feature development.

Game Architect Skill

This skill assists in the documentation phase of game project development. It produces a series of technical documents that guide subsequent implementation work.

[!IMPORTANT] This skill focuses on documentation output only. The actual code implementation phase is outside the scope of this skill.

Output Documents

All documents are placed in an architect/ directory (create if needed). The pipeline produces:

requirement.md  --->  technical_design.md  --->  implementation.md

| Document | Purpose | Description | |----------|---------|-------------| | requirement.md | Requirements Analysis | Analyzes and formalizes user requirements | | technical_design.md | Technical Solution Design | Core document - designs system approaches and patterns | | implementation.md | Implementation Plan | Details data structures, algorithms, class designs, and key code |

Workflow

Phase 0: Ask user if need review

  • Question: "Do you want to review the output document after every phase?"
  • Answer: "Yes" or "No"
  • Default: "Yes"
  • Output: User Review Flag

User Review Workflow:

  • Trigger: User explicitly requests "User Review" mode (via Phase 0).
  • Process:
    • After each Phase, pause and present the output to the user.
    • Request user feedback.
    • If feedback is received, iterate on the current phase's output before proceeding to the next phase.

Phase 1: Requirement Analysis

Goal: Analyze user requirements and produce structured documentation.

  • Input: User request + LLM knowledge
  • Output: architect/requirement.md
  • Reference: Read references/requirements.md

Key Tasks:

  1. Extract and clarify user requirements
  2. Build Feature List (technological scope)
  3. Define Domain Models (for core gameplay)
  4. Document Use Cases & User Flows

Phase 2: Technical Design

Goal: Design technical solutions for each system. This is the most critical phase.

  • Input: architect/requirement.md
  • Output: architect/technical_design.md
  • References:
    • Read references/macro-design.md for high-level structure
    • Read references/principles.md for core principles

Key Tasks:

  1. If existing project: Analyze the exist project code to understand the current architecture
  2. Define Multi-Application structure (Client/Server)
  3. Select Technology Stack (Engine, Languages)
  4. Choose architectural paradigms using the Paradigm Selection Guide for each module
  5. Use the System-Specific References to design each module

Paradigm Selection Guide

| System Type | Recommended Paradigm | Reference | |-------------|---------------------|-----------| | Complex Core Gameplay (Combat, Physics, AI) | Domain-Driven Design (DDD) | references/domain-driven-design.md | | UI, Data Management, Simple CRUD | Data-Driven Design | references/data-driven-design.md | | Rapid Prototyping | Use-Case Driven Prototype | references/prototype-design.md |

System-Specific References

| System Category | Reference | |----------------|-----------| | Foundation & Core (Logs, Timers, Modules, Events, Resources, Audio, Input) | references/system-foundation.md | | Time & Logic Flow (Update Loops, Async, FSM, Command Queues, Controllers) | references/system-time.md | | Combat & Scene (Scene Graphs, Spatial Partitioning, ECS/EC, Loading) | references/system-scene.md | | UI & Modules (Modules Management, MVC/MVP/MVVM, UI Management, Data Binding, Reactive) | references/system-ui.md | | Algorithm & Data Structures (Pathfinding, Search, Physics, Generic Solver) | references/algorithm.md |

Mixing Paradigms

Most projects mix paradigms:

  1. Macro Consistency: All modules follow the same Module Management Framework
  2. Domain for Core: Use DDD for Battle/Core modules
  3. Data for Shell: Use Data-Driven for UI/Meta-game modules
  4. Integration: Application Layer bridges different paradigms

Note

DO NOT contain any Code Snippets in the Technical Design documents.

Phase 3: Implementation Planning

Goal: Create detailed implementation specifications.

  • Input: architect/technical_design.md
  • Output: architect/implementation.md
  • References: Use Specific System Architecture documents from references/

Key Tasks:

  1. Directory Structure: Define the directory structure for the project
  2. Data Structures: Define all core data types and structures
  3. Algorithms: Specify key algorithms with pseudocode
  4. Class Design: Document class hierarchies and relationships
  5. Object Relationships: Define associations, dependencies, and lifecycles
  6. Key Code Snippets: Provide critical implementation examples

Phase 4: Plan Refactoring

Goal: Review and refine the implementation plan for better extensibility and maintainability.

  • Input & Output: architect/implementation.md (in-place update)
  • Reference:
    • Read references/evolution.md
    • Read references/performance-optimization.md (Only if user requires performance optimization)

Refactoring Focus:

  1. Isolation: Ensure proper separation of concerns
  2. Abstraction: Apply appropriate interface abstractions
  3. Composition: Prefer composition over inheritance where applicable
  4. Future Changes: Anticipate and plan for likely evolution

Phase 5: Implementation (Out of Scope)

The final architect/implementation.md is used for actual code implementation.

[!NOTE] Code implementation is not part of this skill. Hand off implementation.md to the implementation phase.

Extensions

1. Refactor Phase (On-Demand)

  • Trigger:
    • "User Review" flag is active.
    • OR User requests a refactor/update for a specific document after the fact.
  • Process:
    • Can target any specific Phase 1 - 5 individually.
    • Input: The existing Output file of that phase (e.g., architect/technical_design.md if refactoring Phase 2). Crucial: Read the file first as the user may have modified it.
    • Goal: Optimize, correct, or expand the document based on specific user feedback or new insights.
    • Output: Update the target file in-place.
    • Note: Phase 4 is a specialized version of this, but the Refactor Phase extension applies generally to any step.

Example Workflows

Example 1: New Formal Project (Core Gameplay Focus)

  • User Input: "I want to start a new ARPG project. The core combat is very complex with many states and interactions. How should I begin?"
  • Execution Path:
    1. Requirement Analysis:
      • Read references/requirements.md.
      • Focus on Domain Model Analysis to capture complex combat concepts.
      • Output: architect/requirement.md.
    2. Technical Design:
      • Read references/macro-design.md. Choose Unity/Unreal and define the layer structure.
      • Read references/principles.md.
      • Select Domain-Driven Design (DDD). Read references/domain-driven-design.md.
      • For Combat Actor structure (EC/ECS), read references/system-scene.md.
      • For Player State Machine (HFSM), read references/system-time.md.
      • For AI pathfinding or spatial queries, read references/algorithm.md.
      • Implement Combat using Entities (Player, Enemy) and Services (DamageCalc).
      • Output: architect/technical_design.md.
    3. Implementation Planning:
      • Output: architect/implementation.md.
    4. Plan Refactoring:
      • Read references/evolution.md.
      • Apply Composition and Abstraction patterns.
      • Update: architect/implementation.md.

Example 2: Adding a UI Module (Data-heavy Focus)

  • User Input: "I need to add a complex Inventory and Shop system to my existing game. How should I design the logic?"
  • Execution Path:
    1. Requirement Analysis:
      • Read references/requirements.md.
      • Focus on Use Cases & User Flow for inventory interactions.
      • Output: architect/requirement.md.
    2. Technical Design:
      • Analyze the exist project code to understand the current architecture.
      • Read references/principles.md.
      • Select Data-Driven Design.
      • Read references/data-driven-design.md.
      • For MVVM and UI Management implementation, read references/system-ui.md.
      • For Resource Caching (Icons/Models), read references/system-foundation.md.
      • Design Item structures, Config tables, and the Global Container for the inventory state.
      • Output: architect/technical_design.md.
    3. Implementation Planning:
      • Output: architect/implementation.md.
    4. Plan Refactoring:
      • Read references/evolution.md.
      • Update: architect/implementation.md.

Example 3: Rapid Prototype

  • User Input: "I have an idea for a unique puzzle mechanic. I want to build a quick demo this weekend to see if it's fun."
  • Execution Path:
    1. Requirement Analysis:
      • Minimal analysis, focus on core puzzle mechanic.
      • Output: architect/requirement.md (lightweight).
    2. Technical Design:
      • Read references/principles.md.
      • Jump to Use-Case Driven Prototype Design.
      • Read references/prototype-design.md.
      • For quick FSM or Update logic, read references/system-time.md.
      • For quick UI (IMGUI), read references/system-ui.md.
      • For core puzzle algorithms (Search/Graph), read references/algorithm.md.
      • Output: architect/technical_design.md (lightweight).
    3. Implementation Planning:
      • Focus on rapid implementation of the core puzzle use case.
      • Output: architect/implementation.md.
    4. Plan Refactoring:
      • Read references/evolution.md.
      • Plan to extract PuzzleController after core mechanic is proven fun.
      • Update: architect/implementation.md.

Example 4: Designing a New System in an Existing Project

  • User Input: "I want to add a Skill System to my current combat engine. It needs to support various effects, cooldowns, and resources. How should I architect it?"
  • Execution Path:
    1. Requirement Analysis:
      • Read references/requirements.md.
      • Focus on Domain Model Analysis to define entities like Skill, Effect, and Requirement.
      • Output: architect/requirement.md.
    2. Technical Design:
      • Analyze the exist project code to understand the current architecture.
      • Read references/principles.md.
      • Apply Domain-Driven Design (DDD) for the core logic (e.g., SkillExecutionService). Read references/domain-driven-design.md.
      • Apply Data-Driven Design for skill configurations (XML/JSON/Excel). Read references/data-driven-design.md.
      • For Scene and Actions, read references/system-scene.md and references/system-time.md.
      • For Event triggering (e.g., OnSkillCast), read references/system-foundation.md.
      • Output: architect/technical_design.md.
    3. Implementation Planning:
      • Output: architect/implementation.md.
    4. Plan Refactoring:
      • Read references/evolution.md.
      • Use Composition (Component pattern) to build complex skills from reusable effects.
      • Use Abstraction (Interfaces) to handle different targeting systems (e.g., Point vs. Target).
      • Update: architect/implementation.md.

Example 5: Architecture and Performance Refactoring

  • User Input: "I've drafted the plan for the new system. Can you review and refactor it? I want to make sure the architecture is clean and scalable, and that it runs efficiently."
  • Execution Path:
    1. Analyze Existing Plan: Review architect/implementation.md to identify coupling issues and potential hotspots.
    2. Read References:
      • Read references/evolution.md to guide architectural separation and extensibility.
      • Read references/performance-optimization.md to find opportunities for caching, pooling, or algorithmic improvements.
    3. Plan Refactoring:
      • Apply Composition to decouple monolithic classes (from evolution.md).
      • Introduce Object Pooling for frequently created entities (from performance-optimization.md).
      • Implement Throttling/Time-Slicing for heavy update loops.
      • Update: architect/implementation.md.