Skills for AI coding agents to master the vvvv gamma ecosystem.
This repository contains agent skills — structured knowledge packages that AI coding agents load automatically when assisting with vvvv gamma development. Each skill covers a specific topic (Spreads, custom nodes, shaders, etc.) and follows the open Agent Skills standard.
Skills work with any compatible coding AI agent, including Claude Code, OpenAI Codex CLI, Cursor, Windsurf, GitHub Copilot, and many others.
vvvv gamma is a live programming environment for .NET — programs run continuously while you build them. C# source files are compiled by vvvv itself via Roslyn and take effect without restart. The ecosystem spans 3D/2D rendering, hardware control (DMX, lasers, depth cameras, robotics), networking (OSC, MQTT, WebSocket), computer vision, audio, and the full .NET NuGet package catalog.
npx skills add tebjan/vvvv-skillsClone the repo and copy the skills/ contents into your agent's skills directory:
# Claude Code (Windows)
git clone https://github.com/tebjan/vvvv-skills %TEMP%\vvvv-skills
xcopy /E /I %TEMP%\vvvv-skills\skills\* %USERPROFILE%\.claude\skills\
# Claude Code (Mac/Linux)
git clone https://github.com/tebjan/vvvv-skills /tmp/vvvv-skills
cp -r /tmp/vvvv-skills/skills/* ~/.claude/skills/Each skill folder (e.g. vvvv-fundamentals/, vvvv-shaders/) must be placed directly inside ~/.claude/skills/ so that each SKILL.md is one level deep. You can also copy skills into a project-local .claude/skills/ directory for per-project scope.
| Skill | Description |
|---|---|
vvvv-fundamentals |
Core concepts — data types, execution model, pads, links, node browser |
vvvv-spreads |
Spread<T> and SpreadBuilder<T> — iteration, mapping, filtering, zipping, Span conversion |
vvvv-custom-nodes |
Writing C# node classes — ProcessNode, Update(), pins, change detection, NodeContext services |
vvvv-patching |
Visual programming — dataflow patterns, regions, patch organization |
vvvv-dotnet |
.NET integration — NuGet packages, .csproj config, vector type interop, async patterns |
vvvv-shaders |
SDSL shaders — TextureFX, shader mixins, compute shaders, ShaderFX composition |
vvvv-node-libraries |
Library project setup — AssemblyInitializer, service registration, ImportAsIs config, node factories |
vvvv-channels |
Public channels — IChannelHub, [CanBePublished], hierarchical propagation, subscriptions, bang channels |
vvvv-editor-extensions |
Editor plugins — .HDE.vl naming, Command nodes, SkiaWindow types, docking, Session API |
vvvv-troubleshooting |
Error diagnosis — C# node issues, shader compilation failures, runtime problems |
Once installed, skills activate automatically when the AI agent detects relevant context in your vvvv gamma project. The agent uses each skill's description field to decide when to load it — no manual invocation needed.
In Claude Code, each installed skill also becomes available as a slash command (e.g. /vvvv-shaders, /vvvv-custom-nodes) for explicit activation.
Example tasks the agent can help with:
- "How do I process a Spread with LINQ?"
- "Create a stateful counter node"
- "Write an SDSL shader that blends two textures"
- "Set up a new node library project"
When the project uses source project references (.csproj), vvvv's live compilation enables a tight feedback loop with AI agents:
- Agent writes .cs files — ProcessNode classes, static methods, shader files
- vvvv auto-recompiles — on file save, Roslyn compiles source into in-memory assemblies
- Nodes hot-swap — updated nodes replace old ones in the running program
- User sees results live — visual output, 3D scenes, hardware signals update in real-time
No build step is needed for source project references — vvvv handles compilation internally. For projects using pre-compiled DLLs, the agent runs dotnet build and the user restarts vvvv. Either way, small modular nodes (each a single C# class) compose into complex systems through visual patching, making it practical to build and iterate on individual pieces with an AI agent.
-
Create a directory under
skills/with a descriptive name (lowercase, hyphens) -
Add a
SKILL.mdwith YAML frontmatter:--- name: vvvv-your-topic description: What this skill does and when to use it. Be specific — agents use this to decide when to activate. license: CC-BY-SA-4.0 compatibility: Designed for coding AI agents assisting with vvvv gamma development metadata: author: Your Name version: "1.0" --- Your instructions here (under 500 lines).
-
Add supporting files (
examples.md,reference.md,templates/) as needed -
Reference supporting files from
SKILL.mdso the agent knows when to load them
- One topic per skill — focused skills compose better than large ones
- Keep SKILL.md under 500 lines — move details to supporting files
- Progressive disclosure — metadata (name + description) is always loaded; SKILL.md body loads on activation; supporting files load on demand. Put "when to use" info in
description, not the body. - Don't explain general programming — AI agents already know C#, .NET, LINQ. Only add vvvv-specific knowledge.
- Write descriptions in third person — "Helps write..." not "I help you..."
- Test with real tasks — ask an AI agent to do vvvv work and verify the skill activates correctly
See Agent Skills Specification for the full format reference.
skills/your-skill/
├── SKILL.md # Required — frontmatter + instructions
├── examples.md # Optional — code examples (loaded on demand)
├── reference.md # Optional — detailed reference (loaded on demand)
└── templates/ # Optional — files the agent can copy/adapt
└── template.cs
- Agent Skills Specification — SKILL.md format, frontmatter fields, progressive disclosure, directory structure
- Agent Skills Documentation Index — complete docs sitemap (fetch this for full reference)
- skills.sh — Open Agent Skills Ecosystem — browse, install, and publish skills
- skills-ref Validation Library — validate SKILL.md frontmatter and structure (
skills-ref validate ./my-skill)
If you use Claude Code, the built-in /skill-creator command can scaffold and validate new skills interactively.
CC BY-SA 4.0 — see LICENSE for details.