LabscoMCP early access ↗
REPO PACKAGECOMMUNITY

vercel-labs ralph-gpu

vercel-labs · publisher58 GitHub starsListed as freegithub.com/vercel-labs/ralph-gpu

LABSCO SUMMARY

Summary

vercel-labs' ralph-gpu package contains exactly one skill, teaching an agent the API of the WebGPU shader library that lives in the same repository.

The skill mirrors the library's own README: initializing a GPU context against a canvas, writing fullscreen fragment passes in WGSL, adding reactive uniforms that update the GPU without reinitializing anything, building compute shaders backed by storage buffers for particle systems, and chaining render targets or ping-pong buffer pairs for iterative effects like blur or a fluid simulation. Every shader an agent writes with it gets automatic access to resolution, time, deltaTime, frame, and aspect uniforms without wiring them by hand.

This is for someone writing WebGPU shader code directly — creative coders building particle systems, fluid sims, or generative visuals in the browser — not for a general app that merely displays a map or a chart. ralph-gpu itself is a small, roughly 11.5kB gzipped, client-side library with no server, account, or MCP dependency: the skill's only real prerequisite is installing the npm package.

READ MORE

There is only the one skill, and it does not go beyond the library's own API.

A package with a spread of task-specific skills would split this into, say, a skill for diagnosing a lost WebGPU context and another for choosing a blend mode for a specific visual effect. Here everything an agent gets is a single reference skill covering the whole surface at once — the entire API in one file, but no separate troubleshooting or use-case guidance layered on top of it.

vercel-labs/ralph-gpu — Repository | Labsco