agentleFS
Sign inStart free
Practice019Practice019/agent-skills/3d-creation/SKILL.md

3d-creation

Skill0 starsChanged 2 months agoMITView on GitHub

3D 创作技能，覆盖两类任务：(1) 用 Blender MCP 从零建模 3D 场景/物体/特效（几何、材质、动画、粒子、物理模拟）；(2) 构建带 3D 元素的叙事性艺术网站（three.js / GSAP / ScrollTrigger / shader，awwwards SOTD 级视觉标准）。只要用户提到 Blender、建模、3D 场景、3D 特效、粒子系统、three.js、WebGL、滚动叙事网站、艺术网站、酷炫官网、awwwards 风格，甚至只是"做一个有 3D 效果的页面"，都应使用本技能。 3D creation skill covering two task types — (1) model 3D scenes, objects, and effects from scratch in Blender via MCP (geometry, materials, animation, particles, physics simulation); (2) build narrative art websites with 3D elements (three.js / GSAP / ScrollTrigger / shader, awwwards SOTD-tier visual standards). Use whenever the user mentions Blender, modeling, 3D scenes, 3D effects, particle systems, three.js, WebGL, scroll-driven narrative sites, art websites, award-tier landing pages, or even just "make a page with a 3D effect".

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the task mixes both (e.g.

    line 18
  2. Distinctive:

    If the task is Blender modeling / VFX and Blender MCP tools are not available, say so and stop — do not pretend the scene was built.

    line 20

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

3D Creation

DSH runtime note: DSH ships no Blender MCP server. If the Blender MCP tools are absent, say so and stop — never pretend a scene was built. The website workflow does not need Blender: build with three.js/GSAP locally, then verify the rendered result with browser-harness screenshots instead of trusting the code.

This skill covers two distinct workflows. Decide which one applies, then read the matching reference file before starting work:

Task typeSignalsRead
Blender 3D modeling / VFXBlender MCP available; user wants a 3D scene, object, animation, particle/physics effect built in Blenderreferences/blender-modeling.md
3D narrative / art websiteUser wants a website with 3D visuals, scroll-driven storytelling, WebGL/shader effects, award-tier aestheticsreferences/narrative-website.md, then references/design-principles.md

If the task mixes both (e.g. "model it in Blender then show it on a website"), read all three references.

If the task is Blender modeling / VFX and Blender MCP tools are not available, say so and stop — do not pretend the scene was built. Website tasks do not require Blender MCP.

The one principle shared by both workflows

These outputs almost never come out right on the first try, and you cannot judge them from code alone. You must render/screenshot the actual result, inspect it visually with your image-understanding capability, diff it against the spec, fix, and look again. Budget multiple iterations from the start:

  • Blender modeling: at least 3 iterations; complex tasks at least 5.
  • Websites: hero screen typically 3–5 rounds; a full page 8–15 rounds is normal.

"The agent said it did X" but the render/screenshot doesn't show it is dishonest design. A report ≠ the actual thing. "No console errors, resources 200" is the floor, not the finish line.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free