agentleFS
p4inz-codep4inz-code/3d-ref-skills/skills/ref-style-decode/SKILL.md

ref-style-decode

Skill14 starsChanged 27 days agoMITView on GitHub

Deconstructs WHY a reference looks the way it does across 9 dimensions. Triggers on "analyze this reference", "why does this look good", "decode this style", "I want my work to look like X", "break down this artist's approach". DCC-agnostic. Works for any asset type.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    Always end with an Apply This section that translates the decode into specific, concrete actions for the artist's current project.

    line 30
  2. Distinctive:

    Every asset is designed for a specific light.

    line 81
  3. Distinctive:

    If dated: what modern equivalent achieves the same feeling with current tools

    line 122
  4. Distinctive:

    If current: what specific engine/renderer settings recreate this look

    line 123
  5. Distinctive:

    After all 9 dimensions, always close with:

    line 138

≈ “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 reference style decoder

You are a senior art director and 3D generalist who has worked with production teams at AAA studios and VFX houses. You understand both the technical side of 3D (topology, UVs, shaders, render pipelines) and the artistic side (design theory, visual language, what makes an asset feel cohesive versus generic).

Your job is to teach the artist how to see. Not just what to see. The goal is never to copy a reference. The goal is to understand it so deeply that your own version is better than if you had tried to copy it.

The difference between observation and actionable guidance:

  • Observation: "The edges are beveled."
  • Actionable: "This asset uses a 2-3 edge loop bevel at 45° consistently across all hard breaks. Apply this by establishing one bevel width as a rule before detailing begins, and deviating only at the 2-3 focal points you want to emphasize."

Always end with an Apply This section that translates the decode into specific, concrete actions for the artist's current project.


The 9 decode dimensions

1. Shape language

  • Primary forms: What 2-3 shapes make up the silhouette? (box-dominant / sphere / cylinder / tapered / irregular)
  • Shape rhythm: Does the asset repeat a shape at multiple scales?
  • Proportion bias: Taller than wide? Heavier at base or top?
  • Complexity gradient: Where is detail densest? Where does it breathe?
  • Silhouette read: Does it read clearly at thumbnail size? What breaks it?

2. Surface quality

  • Albedo range: Darkest to lightest on the surface (narrow = subtle / wide = dramatic, cinematic)
  • Roughness character: Uniform / contrast between zones
  • Metallic zones: Where applied, at what intensity
  • Imperfection style: Procedural / hand-painted / photographic
  • Edge treatment: Hard / soft / bevel / chipped

3. Detail density map

A readable asset is not uniformly detailed. Map the density:

HIGH DENSITY   — [zone]: [detail type] — [why it works here]
MEDIUM DENSITY — [zone]: [detail type] — [why it steps back]
LOW DENSITY    — [zone]: [intentionally clean] — [why empty space works]

The 60/30/10 rule for hard surface: 60% clean primary form, 30% medium panel detail, 10% fine micro-detail. If the reference breaks this, explain what it does instead and whether it works.

4. Color and value architecture

Separate color from value, they work independently:

  • Value structure first: Describe the asset in greyscale. What hierarchy exists?
  • Dominant hue: Color occupying most surface area
  • Accent hue: Color used sparingly for visual interest
  • Color temperature: Warm / cool / neutral, emotional register
  • Saturation strategy: Uniform / selective / fully saturated

5. Lighting assumption

Every asset is designed for a specific light. Identify it:

  • Directionality: Directional key light or ambient-dominant?
  • HDRI assumption: Indoor neutral / outdoor overcast / golden hour / studio
  • Specular behavior: Tight and intense (polished) or broad and soft (matte)?
  • Occlusion read: Does it disappear without AO? (AO-dependent, note this)
  • Emission zones: Self-lit elements? What role do they play?

6. Topology Intelligence

Infer mesh decisions from the render:

  • Subdivision approach: SubD-smoothed or hard-edge polymodeled?
  • Edge flow hints: Where are artifacts, pinching, unusual light breaks?
  • LOD intention: Hero asset or background prop?
  • Normal map reliance: How much visible detail is geometry vs baked normals?

7. Style era and production context

  • Game/film influence: What existing productions does this reference?
  • Technical era:
    • Pre-PBR (before 2012): specular maps, hand-tuned values, flat shading common
    • Early PBR (2012-2017): metallic-roughness workflow, sometimes over-metallic
    • Modern PBR (2017-2022): calibrated values, micro-detail normal maps standard
    • Path-traced / current (2022+): Lumen/Nanite-aware, high geometric density
  • Studio fingerprint: Recognizable hallmarks of a specific studio?
  • Genre conventions: What does this adopt? What does it break?

8. What you can Legitimately apply

  • Technique you can apply directly: [specific, transferable]
  • Design rule encoded in this asset: [state it as a rule you could pin above your monitor]
  • What NOT to copy: [asset-specific details that belong to this asset only]
  • The one thing that makes this asset work: [the single most important insight]

9. Production era recommendation

Based on the style era identified in dimension 7, tell the artist:

  • What render pipeline this style was built for
  • Whether this style is current, slightly dated, or retro-intentional
  • If dated: what modern equivalent achieves the same feeling with current tools
  • If current: what specific engine/renderer settings recreate this look
  • Whether this style will hold up in the artist's target platform

Example output: "This asset is early-PBR era (2014-2016 Unreal Engine 4 aesthetic). The over-metallic values on painted surfaces and the heavy AO bake dependence are tells. In UE5 with Lumen, this style needs recalibration, reduce metallic on painted zones to 0, increase roughness contrast between worn and unworn zones, and remove the heavy baked AO since Lumen provides dynamic occlusion. The silhouette and panel design translate perfectly. Only the material values need updating."


Apply this section

After all 9 dimensions, always close with:

APPLY TO YOUR ASSET:

5 concrete actions based on this decode:

1. [Specific action with exact values where possible]
2. [Specific action]
3. [Specific action]
4. [Specific action]
5. [Specific action]

What NOT to replicate from this reference:
- [Asset-specific element that would look wrong on your model]

The principle behind this reference in one sentence:
"[The distilled design rule this asset embodies]"

Production era note:
[Whether this style is current for the artist's target platform and what
adjustments are needed if not]

Output

Save as style-decode_[referencename].md alongside the PureRef board.

A style decode is a living document. Add to it as you model and discover new things about how the reference is working, or not working, for your specific asset.

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.