← Back to Skills Marketplace
yamasite

Blender 资产生成

by yamasite · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
66
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install blender-asset-generator
Description
Generates ecommerce/game/teaching 3D assets via Blender scripts (Blender 4.2+ including 5.x). Supports cloud Windows and local machines; reference-image guid...
README (SKILL.md)

Blender Asset Generator (Blender 4.2+ / 5.x)

This Skill generates a reproducible asset project that can be executed in:

  • Cloud PC (Windows) as the primary runtime
  • Local computer (Windows or macOS) as an alternative runtime

The Skill outputs:

  • A structured spec (spec.json) describing the asset
  • Executable Blender Python scripts (scripts/*.py) using bpy
  • Run scripts for Windows/macOS
  • Export artifacts: GLB (required), FBX (optional), USDC (optional), plus preview renders

Quick Start

  1. Choose a mode:

    • Text-only Mode (no image or no vision): fast and deterministic
    • Vision-assisted Mode (reference images + multimodal model): extract structure/dimensions from images, then confirm with user
  2. Provide input:

    • Required: object name + use case (ecommerce / game / teaching) + desired output formats (GLB required, FBX optional)
    • Optional: target size (at least one anchor dimension), poly budget, material style, reference images
  3. The Skill will:

    • Normalize input into spec.json
    • Generate scripts/build.py, scripts/render.py, scripts/export.py (or update them)
    • Provide Cloud (Windows) and Local (Windows/macOS) run commands
    • Generate a QA checklist and a final evaluation

Input Template

Blender Version: 4.2+ (supports 4.2 and 5.x)
Primary Runtime: Cloud PC (Windows)
Output Priority: GLB required; FBX optional
USD: USDC optional (via file extension)

Use Case: ecommerce | game | teaching
Asset Name:
Asset Category: (e.g., bottle / lunch box / hanger / simple prop / teaching diagram)

Text Description (required):
- What it is:
- Key parts/components:
- Material style (optional): plastic/metal/glass/wood/fabric
- Target realism: low/medium/high

Dimensions (recommended):
- Provide at least ONE anchor dimension, e.g. height=240mm
- If unknown: provide category-typical size OR accept default size

Constraints (optional):
- Poly budget (triangles):
- UV: required | optional | no
- Texture placeholders: yes | no
- Naming convention: (optional)

Reference Images (optional):
- Attach images (front/side/top) if available
- If images are provided, say whether there is a scale reference in the image:
  - e.g. A4 paper / coin / ruler / hand / known product standard size

References

Read these as needed:

  • Workflow: references/workflow.md
  • Ecommerce reference-image template: references/ecommerce-reference-guided-template.md
  • Degradation strategy (no vision available): references/degradation-strategy.md
  • QA checklist: references/qa-checklist.md
  • Presets (common defaults): references/presets.md

Output Contract

Always output in this structure:

  1. Project Package
    • File tree (what to create/update)
    • spec.json (final)
  2. Run Instructions
    • Cloud PC (Windows) command(s)
    • Local (Windows) command(s)
    • Local (macOS) command(s)
  3. Expected Artifacts
    • scene.blend
    • exports/*.glb (required)
    • exports/*.fbx (optional)
    • renders/*.png (required: at least 2 views)
    • run-log.txt (required)
  4. QA + Evaluation
    • Checklist results
    • Verdict: Achieved | Partially Achieved | Not Achieved
    • Top fixes

Behavior Rules

  • Determinism: prefer parametric geometry and explicit settings; avoid random seeds unless user asks.
  • Evidence: do not invent certifications/performance claims; in 3D context, do not assume hidden dimensions from a single photo.
  • Reference images:
    • If a multimodal model is available, extract a draft structure card and dimension hypotheses, then ask for confirmation of at least one anchor dimension.
    • If vision is NOT available, never block the workflow. Use Text-only Mode + ask for a minimal clarification form (see references/degradation-strategy.md).
  • Units: default to millimeters in the spec; enforce unit conversion in scripts.
  • Exports: always export GLB; export FBX only when requested.

Assets (Project Skeleton)

Use the project skeleton to generate a runnable folder:

  • assets/project-skeleton/spec.schema.json
  • assets/project-skeleton/spec.example.json
  • assets/project-skeleton/scripts/build.py
  • assets/project-skeleton/scripts/render.py
  • assets/project-skeleton/scripts/export.py
  • assets/project-skeleton/run_local_windows.ps1
  • assets/project-skeleton/run_local_macos.sh
  • assets/project-skeleton/run_cloud_windows.ps1
Usage Guidance
This package appears coherent for generating Blender 3D assets. Before running: (1) ensure you have a compatible Blender binary installed and point BLENDER_BIN if needed; (2) review the included Python scripts yourself (they are plain bpy scripts that read spec.json and write files) and run them in a controlled environment; (3) if you plan to run this on a cloud PC, confirm the cloud runtime permits executing Blender and storing outputs there; (4) note that 'vision-assisted' behavior depends on a vision-capable agent/runtime — the skill does not itself access cameras or external servers. If any of those are unacceptable, do not run the scripts.
Capability Tags
crypto
Capability Assessment
Purpose & Capability
Name/description (Blender asset generation) aligns with the included artifacts: Blender Python scripts (build/render/export/donut_build), run scripts for macOS/Windows, spec.json/schema, and reference docs. There are no unrelated environment variables, binaries, or config paths requested.
Instruction Scope
SKILL.md and the referenced docs limit actions to normal asset-generation tasks: normalize inputs to spec.json, produce Blender scripts, run Blender in background mode, generate renders and exports, and produce logs. The documents mention optional 'vision-assisted' steps but do not include any code that accesses camera hardware, system secrets, or external endpoints; vision use is conditional on a vision-capable agent/environment.
Install Mechanism
There is no install spec and no network downloads; the skill ships a project skeleton and Blender scripts. All actions write local project files and call the Blender binary — no external installers or extract-from-URL steps are present.
Credentials
The skill requests no credentials or special environment variables. It reads/writes files within the project folder (spec.json, scene.blend, exports, renders, run-log.txt) which is expected for this purpose. No secret-named env vars or unrelated service tokens are requested.
Persistence & Privilege
always is false and autonomous invocation is not disabled (the default); this is normal for skills. The skill does not attempt to modify other skills or global agent config; its footprint is limited to project files it creates/updates.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install blender-asset-generator
  3. After installation, invoke the skill by name or use /blender-asset-generator
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of blender-asset-generator. - Generates structured, reproducible 3D asset projects for ecommerce, game, or teaching use cases via Blender scripts (supports Blender 4.2+ and 5.x). - Supports cloud Windows PC as primary runtime, with local execution on Windows or macOS as alternatives. - Offers text-only and vision-assisted workflows; gracefully degrades if reference images or vision are unavailable. - Produces clear project structure: spec.json, Blender Python scripts, run scripts, and required export formats (GLB mandatory, FBX/USDC optional), plus preview renders and QA evaluation. - Includes input templates and references to guide usage and ensure consistent quality.
Metadata
Slug blender-asset-generator
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Blender 资产生成?

Generates ecommerce/game/teaching 3D assets via Blender scripts (Blender 4.2+ including 5.x). Supports cloud Windows and local machines; reference-image guid... It is an AI Agent Skill for Claude Code / OpenClaw, with 66 downloads so far.

How do I install Blender 资产生成?

Run "/install blender-asset-generator" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Blender 资产生成 free?

Yes, Blender 资产生成 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Blender 资产生成 support?

Blender 资产生成 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Blender 资产生成?

It is built and maintained by yamasite (@yamasite); the current version is v1.0.0.

💬 Comments