agentleFS
EL4CTEOEL4CTEO/blender-skills/skills/blender-scene-setup/SKILL.md

blender-scene-setup

Skill779 starsChanged 14 days agoMITView on GitHub

Set up and audit Blender scenes the way studios do - units and real-world scale, collection layout, naming conventions (SM_/SK_/T_/M_), origins, linking vs appending, library overrides, asset libraries and catalogs, collection exporters, orphan/path cleanup. Use when starting a project, organizing a .blend, or before handoff/export.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run scripts/scene_audit.py (add --fix purge scale paths only on files you own).

    line 15
  2. Distinctive:

    Use Collection Exporters (4.2+) to define the export once per collection, then re-export with one call:

    line 32
  3. Distinctive:

    Use ASCII only, no spaces, and PascalCase or snake_case consistently across the project.

    line 52
  4. Very specific:

    Keep library paths relative (//).

    line 60
  5. Very specific:

    Import method per library: LINK, APPEND, APPEND_REUSE (default.

    line 66
  6. Distinctive:

    Never ship multi-user mesh data that has an unapplied per-object scale.

    line 71

≈ “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

Scene Setup & Hygiene

The execution model and verification steps are in blender-core.

Procedure

  1. Run scripts/scene_audit.py (add --fix purge scale paths only on files you own). It reports units, unapplied/negative scale, default names, objects in the scene root, missing images, orphans and far-off origins.
  2. Set units: Metric, Unit Scale 1.0, Length meters. Keep 1.0 even for Unreal/Roblox, because exporters handle the conversion (see blender-export-pipelines).
  3. Build the collection layout (below). Every object must be in a named collection, never the scene root.
  4. Name everything with the conventions below. Rename after every duplicate or join so no .001 suffixes remain.
  5. Set origins: props at the bottom-center (they sit on the ground at Z=0), characters at world origin between the feet, doors/hinges at the pivot.
  6. Apply rotation and scale on meshes (Ctrl+A). Keep location unless the asset should export at the origin.
  7. Re-run the audit. It's done when issues is empty or every remaining issue is intentional and explained.

Collection layout (one asset per .blend, or one collection per asset)

ASSET_Crate
├── Crate_high      # sculpt / subdiv source (exclude from export)
├── Crate_low       # game mesh(es) SM_Crate, SM_Crate_LOD1..
├── Crate_bake      # cages, bake helpers
├── Crate_collision # UCX_SM_Crate_00 (UE) / Crate-col (Godot)
└── Crate_ref       # reference images/blockout (hide_render)

Use Collection Exporters (4.2+) to define the export once per collection, then re-export with one call:

with bpy.context.temp_override(collection=col):
    bpy.ops.collection.exporter_add(name="IO_FH_gltf2")
ex = col.exporters[-1]; ex.export_properties.filepath = "//export/SM_Crate.glb"
with bpy.context.temp_override(collection=col):
    bpy.ops.collection.exporter_export(index=len(col.exporters) - 1)

bpy.ops.wm.collection_export_all() re-runs every collection exporter.

Naming conventions (Unreal-style, accepted by all engines)

TypePatternExample
Static meshSM_<Name>[_LOD#]SM_Crate_LOD1
Skeletal mesh / rigSK_<Name> / RIG_<Name> (armature object Armature or root)SK_Hero
MaterialM_<Name> / instance MI_<Name>M_Wood_Oak
TextureT_<Name>_<Suffix>: _BC base color, _N normal, _ORM (R=AO G=Rough B=Metal), _E emissive, _M maskT_Crate_ORM
CollisionUE UCX_<Mesh>_##, Godot suffix -col/-colonly/-convcolonlyUCX_SM_Crate_00
Bones/sides.L/.R suffix (Blender mirror tools depend on it)upper_arm.L
Actions<Rig>_<Clip>Hero_Run
Use ASCII only, no spaces, and PascalCase or snake_case consistently across the project.

Real-world scale reference (meters)

Door 2.0–2.1 × 0.9 · ceiling 2.4–2.7 · table 0.75 · chair seat 0.45 · counter 0.9 · stair rise 0.18 / tread 0.28 · adult human 1.7–1.8 (eyes 1.6) · car 4.5 × 1.8 × 1.5. Roblox: 1 stud ≈ 0.28 m; default character ≈ 5 studs tall. Wrong scale breaks physics, lighting falloff, DOF and texel density, so check ob.dimensions first.

Linking, appending, overrides

  • Append (libraries.load(link=False)) makes a local copy. Use it for one-off reuse.
  • Link (link=True) is read-only, updates from the source and is what pipelines use. To edit a linked collection, add a library override: ob.override_hierarchy_create(scene, view_layer, do_fully_editable=False). Only overridable properties (transforms, pose, and anything marked overridable) can change.
  • Keep library paths relative (//). Check bpy.data.libraries for broken paths.

Asset libraries

  • Mark: ob.asset_mark(); ob.asset_generate_preview(). Metadata goes in ob.asset_data (author, description, license, tags.new("wood"), catalog_id).
  • Catalogs live in blender_assets.cats.txt at the library root, one line per catalog: UUID:Path/Sub:Simple Name.
  • Register a library with prefs.filepaths.asset_libraries.new(name=..., directory=...). 5.2+: remote libraries via use_remote_url + remote_url. In 5.2 active_asset_library indices are offset by 2 ("All", "Essentials").
  • Import method per library: LINK, APPEND, APPEND_REUSE (default; reuses already-appended data).

Hygiene rules

  • Purge orphans before shipping (bpy.data.orphans_purge(do_recursive=True)). Use fake users only on purpose.
  • Image paths: relative (file.make_paths_relative), or pack for single-file delivery (file.pack_all).
  • Never ship multi-user mesh data that has an unapplied per-object scale. Instances are fine, but bake scale before export.
  • Hide helpers with hide_render=True and exclude them from export collections; viewport hiding alone isn't enough.
  • One active scene per file for delivery. Extra scenes confuse exporters (use_active_scene).

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.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.