AI agent reminder

Unity skills

Use Unity Technologies’ reusable AI agent skills when an agent is helping with project setup, packages, UI, URP, builds, or multiplayer services.

Open GitHub repo Project uses

Install

When working in a local AI coding agent that supports Agent Skills, install Unity’s skill pack with:

npx skills add Unity-Technologies/skills

Reminder: install this in the environment where the AI agent is running, not on the public web server unless the Unity project is actually being developed there.

Best skills for this project

`unity-cli`

Use for Unity editor installs, project checks, command-line builds, WebGL builds, batch mode tasks, and general project automation.

`unity-package-management`

Use when adding or reviewing packages for URP, Input System, Cinemachine, UI Toolkit, Netcode, Multiplayer Services, or testing tools.

`ui`, `ui-uitk`, `ui-ugui`

Use when building workstation screens, switch front panels, CLI overlays, target-device panels, instructor menus, or lesson UI.

`setup-multiplayer-services`

Use later if the lab becomes multiplayer: instructor/student sessions, shared lab rooms, lobby discovery, or cooperative troubleshooting exercises.

`validate-urp-render-graph-renderer-feature`

Use if custom URP renderer features are added for wireframe cable views, x-ray connection overlays, port highlighting, or training visualization effects.

Prompts to reuse

Good prompts to give an AI agent after installing the Unity skills:

“Use the Unity skills. Inspect my project and summarize the Unity version, render pipeline, packages, scenes, and build target.”
“Use Unity package management to recommend packages for a WebGL URP networking training simulator with UI panels and a CLI.”
“Help me create a data-driven model for devices, endpoints, cables, link state, power state, VLANs, and lesson success conditions.”
“Build a UI prototype for a 24-port switch panel with clickable ports, LED states, connected cable art, and a CLI area.”
“Plan a first vertical slice: one switch, one router, one workstation, one server, and one disconnected-port troubleshooting lesson.”

Development workflow

1. Ask the agent to inspect first

Before code changes, have the agent read project structure, Unity version, packages, scenes, prefabs, scripts, and input setup.

2. Keep changes vertical

Prefer one complete training loop over many half-built systems: select a port, change connection state, update LED, update CLI, verify lesson.

3. Protect WebGL constraints

Ask the agent to consider WebGL performance, build size, lighting cost, input capture, memory, and browser deployment with every feature.

4. Make lessons data-driven

Have the agent create scenario data instead of hardcoding every fault. The instructor should eventually be able to configure new lessons.