niaka3dayo/agent-skills-vrc-udon
Skills, rules, and validation hooks that teach AI coding agents to generate correct UdonSharp code
What it solves
Developing for VRChat using UdonSharp involves strict constraints that differ from standard C#. AI coding agents often struggle with these limitations, frequently suggesting incompatible features like async/await, try/catch, or List<T> that cause compile errors. This project provides the necessary rules and validation to ensure AI-generated code is actually compatible with the VRChat runtime.
How it works
This project acts as a knowledge base and validation layer for AI coding assistants. It provides:
- Rules: Constraint files that guide AI agents on what is allowed in UdonSharp, including networking patterns and sync decision trees.
- Validation Hooks: Automated scripts that check your code after editing to catch common mistakes, such as using blocked features or improper networking configurations.
- Templates: Pre-configured code structures for common VRChat tasks.
Who it’s for
This is designed for VRChat world creators who use AI coding agents (like Claude Code) to help write UdonSharp scripts and want to avoid the frustration of AI-generated code that fails to compile.
Highlights
- Provides specific rules to prevent common UdonSharp compile errors.
- Includes automated validation hooks for Bash and PowerShell to catch issues in real-time.
- Offers a sync decision tree to help optimize network usage and avoid data bloat.
- Features a library of code templates for common interactions and networking patterns.
- Supports the latest VRChat SDK (3.10.6) with clear documentation on feature mapping.
Related
- Project
- Project
- Project
- Project
- Project