vgpu skill
A Vercel Labs skill for the vgpu WebGPU library: WGSL shaders, 3D scenes, tests and Blender assets
- 2,652 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill tells the agent to run the local vgpu CLI through npx and package managers
- The optional hosted vgpu.sh MCP server reaches out over the network
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/vgpuDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill tells the agent to run the local vgpu CLI through npx and package managers
- The optional hosted vgpu.sh MCP server reaches out over the network
This is third-party code. Review the repository files before installing.
What it does
The vgpu skill from Vercel Labs helps an agent write, debug and speed up WebGPU projects with the vgpu library, its CLI and the @vgpu packages. The skill itself is thin: it carries no API reference, it takes documentation from the version of the package installed in the project through `npx vgpu docs` and looks for examples with `npx vgpu examples search`. While local package docs exist, the skill forbids inferring API shapes from memory, repository branches or hosted documentation. Before working, the agent finds out which vgpu version sits in the relevant part of a monorepo and checks the local CLI with `pnpm exec vgpu --version` or `npm exec --no -- vgpu --version`. The `--no` flag stops the package manager from downloading a missing command. For 3D scenes the skill points to the scene.md guide on hierarchies, transforms, instances, cameras and animation and physics integration. A separate section covers Blender: modeling, high and low poly workflows, normal and ambient occlusion baking, levels of detail and asset validation. The guide does not depend on the vgpu version; for runtime API the agent goes to local docs. The library itself writes shaders as typed WGSL modules and runs in the browser, in Node with Dawn and in the `vgpu/mock` test mock.
Who it is for. Graphics programmers and frontend developers on WebGPU who write shaders and scenes with vgpu alongside Claude Code, Cursor or Codex.
Good fit when
- A WGSL shader is imported as a TypeScript module and the agent checks types against the docs of the installed version
- A scene with a camera, transforms and instances is built from scratch following the scene.md guide
- A Blender model is prepared for runtime: normal and ambient occlusion baking, LODs, asset validation
- Graphics tests run in CI without a GPU, and the agent wires up `vgpu/mock`
- The task needs a ready example: the agent finds it in the gallery and pulls the source with `npx vgpu examples pull`
Not a fit when
- A Three.js or Babylon.js project without vgpu: the skill points to the vgpu package docs
- vgpu is not installed in the project: the skill reads docs from the package and has nothing to rely on without it
- You need a general WebGPU course: the skill knows vgpu and does not replace a browser API tutorial
Example request
Build a fullscreen effect on vgpu with a wave shader and time in a uniform, render it on a canvas and add a test on the mock without a GPULimitations
- The skill is version-neutral and contains no API reference. Without the `vgpu` package and its CLI installed it adds little.
- The README marks vgpu as a Vercel Labs product, with no API stability guarantee.
- The hosted MCP server `https://vgpu.sh/api/mcp` reads documentation from the network. To tie to the package version, the README suggests the local `npx vgpu mcp`.
- The Blender section does not depend on the vgpu version, and runtime API comes from local docs.
How to disable. Delete the vgpu skill folder from the agent's skills directory. Your package manager removes the package from the project, for example `pnpm remove vgpu`.
FAQ
How do I install the vgpu skill in Claude Code?
Run `npx skills add vercel-labs/vgpu` in the project directory. The library is installed separately, for example `pnpm add vgpu`, and types need `pnpm add -D @webgpu/types`.
Where does the skill get API documentation?
From the vgpu package installed in the project, through `npx vgpu docs`. The skill is version-neutral, so its hints do not go stale when the library version changes.
Does vgpu have an MCP server?
Yes, in two forms. The hosted read-only public server lives at `https://vgpu.sh/api/mcp`, and the local one runs with `npx vgpu mcp`, bound to the project with `--project-from-cwd`.
What does the skill cover for Blender?
It guides modeling, high and low poly variants, normal and ambient occlusion baking, levels of detail and asset validation. Loading the finished model into the runtime follows the vgpu package docs.
Which runtimes does vgpu support?
The browser, headless Node through `vgpu/node` on Dawn, and a deterministic mock `vgpu/mock` for tests and CI. One public API works in all three.
Discussion
No questions or reviews yet.
Related
SkillsA skill that dissects and ports WebGL, WebGPU and Canvas effects from a page into a standalone local JS project
A skill for building WebGPU apps on Three.js with TSL shaders: renderer setup, node materials, compute shaders, post-processing
Skill and MCP server for Shaders: the agent builds backgrounds, glass, gradients and cursor effects on WebGPU for React, Vue, Svelte, Solid and JS
A skill builds and debugs GLSL shaders from reusable blocks, bundles dependencies into one file and checks WebGL2-oriented code.
Three skills for web 3D: GLSL shaders, Three.js scenes with React Three Fiber and GPU particle systems.
A skill that guides an agent from renderer choice to browser verification for small GPU effects in interfaces on WebGL and WebGPU.
README in short
The README presents vgpu as a TypeScript WebGPU library with typed WGSL imports, a single `Gpu` context and a small size: a fullscreen effect takes 25 KB gzipped. A quick start follows, then a section on agent resources (CLI, skill, MCP, `llms.txt`) and the monorepo's packages. MIT license.
README badge
Original SKILL.md
--- name: vgpu description: >- Build, debug, test, and optimize WebGPU projects using vgpu, its CLI, or @vgpu packages. Use for vgpu API questions, WGSL workflows, browser or Node rendering, integrations, testing, performance work, and Blender asset modeling, baking, and runtime integration. --- # vgpu Treat the documentation bundled with the target project's installed `vgpu` package as the authority for that project. This skill is intentionally version-neutral: do not infer API shapes from the skill's Git revision, remembered APIs, the repository default branch, hosted docs, or a hosted MCP server when local package docs are available. The public skill install command has no branch pin: ```sh npx skills add vercel-labs/vgpu ``` ## 3D scenes If you need to work with 3D scenes, read [Scene composition](./scene.md) before implementing scene hierarchies, transforms, instances, cameras, or animation and physics integration. ## Blender assets For Blender modeling, high/low-poly workflows, normal or ambient occlusion baking, LODs, and asset validation, read [Blender assets](blender/index.md). That guide links to focused references for shape and assembly or baking diagnostics; load them when the task needs them. Its authoring guidance is independent of the vgpu version. For runtime API work, use the package docs below. ## Select the package version Run the project-local `vgpu` executable from the workspace that owns the dependency. First check the relevant package manifest and lockfile, then confirm which local CLI you resolved without allowing the package manager to download a missing command: ```sh pnpm exec vgpu --version npm exec --no -- vgpu --version ``` Use a local-only equivalent for other package managers, such as `yarn exec vgpu --version` or an already resolved `node_modules/.bin/vgpu`. Commands such as bare `npx vgpu` or `bunx vgpu` may download a missing package, so do not use them for local version discovery. npm's `--offline` is also insufficient because it can install a missing package from cache. Do not replace an installed version merely to read newer documentation.