shader-dev
by MiniMax-AIshader-dev is a practical GLSL shader skill for ShaderToy-style real-time visuals. Use the shader-dev skill to build or debug ray marching, SDF scenes, lighting, particles, fluid motion, post-processing, and shader-dev for UI Design with less guesswork than a generic prompt.
This skill scores 84/100, which means it is a solid listing candidate for directory users who want a reusable GLSL shader workflow instead of generic prompting. The repository provides enough routing, technique coverage, and reference depth to help an agent trigger the skill correctly and work with less guesswork, though it is still specialized and lacks external helper files or install-time automation.
- Explicit invocation pattern is documented (`/shader-dev <request>`), making triggerability clear for agents.
- Broad, structured coverage of 36 ShaderToy-compatible GLSL techniques with a routing table and many detailed reference files.
- Substantial workflow content and no placeholder/demo markers; the body includes headings, code fences, repo references, and practical guidance for shader tasks.
- No scripts, resources, or install command are provided, so adoption depends on reading the SKILL.md routing instructions.
- Highly specialized to real-time GLSL/ShaderToy-style shader work, so it is less useful for general coding or non-graphics agents.
Overview of shader-dev skill
What shader-dev is for
shader-dev is a practical GLSL shader skill for turning visual ideas into real-time effects, especially ShaderToy-style work. It helps when you want to build or debug ray-marched scenes, SDF compositions, lighting, procedural textures, particles, fluid motion, post-processing, or camera effects without guessing which shader technique fits best.
Who should use it
This shader-dev skill is best for graphics developers, technical artists, UI engineers exploring motion-rich interfaces, and AI agents asked to produce shader code from a rough visual brief. It is especially useful when the task is “make this effect work in GLSL” rather than “explain the theory.”
What makes it different
The main value of shader-dev is technique routing: it organizes many shader topics into a structured path instead of treating all visuals as one generic prompt. That means the skill can narrow a request like “glowing tunnel with fog and soft shadows” into the right building blocks faster than a one-shot prompt. It is also broad enough to cover both composition and rendering concerns, which matters when the blocker is not just shape generation but also quality issues like aliasing, shadows, and buffer workflow.
How to Use shader-dev skill
Install and trigger it correctly
Install with npx skills add MiniMax-AI/skills --skill shader-dev, then invoke it with a focused request such as /shader-dev create a raymarched SDF crystal scene with rim lighting and soft shadows. The skill is designed to work from the user’s arguments, so the prompt should describe the visual goal, target platform, and constraints instead of only naming a technique.
Give the skill the right input shape
Strong input usually includes:
- the effect type:
ray marching,SDF,fluid,particle system,post-processing - the target environment:
ShaderToy, WebGL, or another GLSL runtime - the visual intent: “clean UI background,” “cinematic fog,” “game-like neon tunnel”
- constraints: performance, mobile support, one-pass vs multipass, no textures, no 3D assets
A weak request like “make it cool” forces the skill to guess. A stronger request like “build a ShaderToy-compatible SDF corridor with animated panels, ambient occlusion, and antialiasing, optimized for real-time use” gives it enough detail to choose the right technique path.
Read the right files first
Start with SKILL.md for invocation and routing, then inspect the most relevant techniques/ file before jumping to a deeper reference/ page. For example:
ray-marching.mdfor sphere tracing scenessdf-3d.mdorsdf-2d.mdfor object constructionlighting-model.mdfor shading choicesprocedural-noise.mdfor organic variationpost-processing.mdfor bloom, tone mapping, or screen-space polish
Use the reference/ files when you need implementation detail, math, or edge-case handling. This keeps the first pass fast and avoids over-reading unrelated shader theory.
Workflow that improves output quality
A good shader-dev usage workflow is: identify the core rendering approach, confirm the target runtime, ask for the main effect first, then layer supporting effects after the base scene works. For example, generate the geometry and camera first, then add shadows, AO, anti-aliasing, and color grading. That order reduces broken shader code and makes debugging easier.
shader-dev skill FAQ
Is shader-dev only for ShaderToy?
No. The skill is ShaderToy-compatible in spirit, but the underlying GLSL ideas can be adapted to other real-time shader environments. If your engine has different uniforms, buffer setup, or texture rules, mention that early so the output fits your runtime.
When should I not use shader-dev?
Skip shader-dev if you need a simple static gradient, a CSS animation, or a non-shader UI effect that does not rely on GLSL. It is most valuable when the final output depends on shader math, rendering pipelines, or real-time visual synthesis.
Is shader-dev better than a generic prompt?
Usually yes, because the shader-dev skill provides technique-aware guidance instead of forcing the model to infer the correct rendering strategy from scratch. The gain is biggest when the request involves multiple interacting systems, like SDF geometry plus lighting plus buffer-based feedback.
Is shader-dev beginner-friendly?
Yes, if you describe the desired look in plain language and let the skill route the technique. Beginners get the best results by starting with one visible goal, such as “make a foggy portal” or “build a glowing UI orb,” rather than asking for an entire production-grade shader in one shot.
How to Improve shader-dev skill
State the visual target before the technique
The best shader-dev guide inputs describe the scene outcome first and the math second. “A slow-moving neon water surface with horizon haze and reflective highlights” is better than only “use domain warping,” because the skill can map the effect to the right combination of techniques.
Add the constraints that usually break shaders
If you want better results, say what must stay true:
- ShaderToy compatibility or not
- single-pass vs multipass
- performance budget
- texture availability
- 2D, 3D, or screen-space output
- whether animation should be subtle, loopable, or interactive
These details matter because many shader failures are not visual failures but pipeline mismatches.
Iterate from a baseline, not from the final polish
For shader-dev for UI Design, ask for the simplest version of the effect first, then refine. A solid improvement loop is: base shape → motion → lighting → antialiasing → color palette → post-processing. If the first output is off, correct the rendering approach or scene structure before asking for stylistic tweaks.
Provide failure feedback in technical terms
If the result is too noisy, too slow, too flat, or too busy, say so directly and connect it to the symptom. For example: “reduce temporal flicker and simplify the SDF detail” is more useful than “make it cleaner.” The skill can then adjust sampling, shape complexity, shading, or color treatment instead of randomly restyling the scene.
