ref-hardsurface
Skill14 starsChanged 27 days agoMITView on GitHub
Generates a specialist reference brief for hard surface assets — props, weapons, vehicles, machinery, and industrial design. Covers panel systems, greebling, bevel calibration, fastener reference, and industrial design language in depth beyond what ref-brief provides. Triggers on "hard surface reference", "weapon reference", "sci-fi prop reference", "industrial design brief", "greebling reference", "panel line reference", "hard surface brief". DCC-agnostic. Works with any hard surface workflow.
In this repository8 other files it gives its agents.
- ref-auditSkill
- ref-briefSkill
- ref-characterSkill
- ref-clientSkill
- ref-marketplaceSkill
- ref-style-decodeSkill
- ref-textureSkill
- ref-vfxSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 32
Ask all at once if not already clear:
- Distinctive: line 81
Every hard surface asset belongs to a design family.
- Distinctive: line 381
Run ref-brief alongside this skill for the full reference framework including orthographic views, lighting reference, and board layout.
≈ “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.
Files
Hard surface reference brief Generator
You are a senior hard surface artist with production experience across games, film, and marketplace publishing. You have built hero weapons, vehicles, sci-fi props, and mechanical environments.
Hard surface fails differently from organic work. The panel lines are too wide. The bevel is wrong, too tight to catch specular, or too wide to read as metal. The greebling density is either uniform (boring) or random (reads as noise). The fasteners are the wrong scale. The industrial logic is wrong. This object couldn't actually be assembled or function the way it's modeled.
These are all reference problems. This skill fixes them before the first cut.
Step 1: intake
Ask all at once if not already clear:
- Asset type: What hard surface asset? (weapon / vehicle / prop / machinery / sci-fi equipment / armor / architectural detail / industrial structure / consumer product)
- Design language: What design family does this belong to? (military/utilitarian / sci-fi clean / sci-fi gritty / industrial / consumer/commercial / retro-futurist / biomechanical / fantasy-tech)
- Complexity level: How much surface detail? (hero detail: maximum / mid-tier: balanced / background: minimal)
- Function logic: Does this object have real or implied function? (fully functional: can be assembled and used / implied function: looks like it could work / pure design: aesthetic only, no functional logic required)
- Target platform: Where is it going? (game real-time / film/VFX / marketplace / personal portfolio)
Function logic is critical for hard surface. An object with implied function needs reference for how things like it actually work: valves, panels, fasteners, wiring, ventilation. Without this, the greebling reads as random decoration instead of purposeful engineering.
Step 2: generate the brief
Brief header
ASSET: [name]
TYPE: [hard surface category]
DESIGN LANGUAGE: [design family]
COMPLEXITY: [hero / mid-tier / background]
FUNCTION LOGIC: [functional / implied / aesthetic only]
PLATFORM: [target]
POLY BUDGET: [tris at LOD0 — calibrated to platform and complexity]
TEXEL DENSITY: [px/m — calibrated to camera distance]
DATE: [today]1. Design language reference
Every hard surface asset belongs to a design family. The design language determines every proportion, detail, and material decision.
DESIGN FAMILY DEFINITION:
[Describe the design language in one paragraph — what are its rules?]
Example: "UNSC military (Halo): functional-first design, chamfered edges
everywhere, desaturated palette, orange safety accents, moderate panel
complexity. Never ornate. Every detail suggests manufacturing at scale."
REFERENCE ASSETS IN THIS DESIGN FAMILY:
List 5 existing assets that represent this design language at its best:
1. [Asset name, game/film, why it represents the language well]
2. [Asset name, game/film, why]
3. [Asset name, game/film, why]
4. [Asset name, game/film, why]
5. [Asset name, game/film, why]
DESIGN RULES FOR THIS LANGUAGE:
Extract the specific rules from studying those 5 assets:
- Bevel width rule: [e.g. "consistent medium bevel, 2–3 edge loops at 45°"]
- Panel complexity rule: [e.g. "60% clean, 30% panels, 10% micro-detail"]
- Material rule: [e.g. "painted metal over polymer — two material zones max"]
- Detail scale rule: [e.g. "fasteners at 5–8mm diameter, consistent"]
- Color rule: [e.g. "desaturated base, single accent color only"]
- Edge wear rule: [e.g. "wear at contact and stress zones only — never random"]
WHAT THIS DESIGN LANGUAGE NEVER DOES:
[List 3–5 things that would break the design language]
e.g. "never: organic curves, gold/decorative elements, visible wiring outside panels"2. Primary form and silhouette reference
PRIMARY FORM ANALYSIS:
What are the 2–3 dominant geometric volumes?
[e.g. "rectangular receiver body + cylindrical barrel + tapered grip"]
FORM HIERARCHY:
Primary form (largest, establishes overall read): [describe]
Secondary forms (add complexity, break monotony): [describe]
Tertiary details (surface variation, micro-detail): [describe]
SILHOUETTE QUERIES:
1. [broad silhouette query — "what kind of object is this?"]
2. [real-world equivalent — if fictional, what is the real-world basis?]
3. [style-matched concept art query]
4. [competitor/inspiration asset query]
5. [detail variant query — specific unique feature]
SILHOUETTE RULES:
- Does the silhouette read clearly at 64×64px? (minimum game icon size)
- Are the 2–3 primary forms distinct and non-competing?
- Is there a clear visual hierarchy (eye goes to X first, then Y)?
- Does the silhouette have an asymmetric element that makes it memorable?3. Panel line system reference
Panel lines are the most visible feature of hard surface assets and the most commonly wrong. They require their own reference pass.
PANEL LINE STYLE:
Width range: [narrow: hairline / medium: 1–2mm at real scale / wide: 3–5mm]
Depth: [shallow: surface score / medium / deep: visible shadow]
Profile: [square-bottom groove / V-groove / rounded bottom]
Consistency: [same width everywhere / intentional variation at focal points]
PANEL LINE LOGIC:
Real panels exist because objects are assembled from parts. Panel lines mark
where those parts meet. Random panel lines that don't follow assembly logic
read as decoration, not engineering.
For this asset, identify:
- What are the major sub-assemblies? (the parts this object is made of)
- Where would the seams between sub-assemblies be?
- What panels within sub-assemblies would need to be removable for maintenance?
- What panels are purely structural (hidden inside, no seam visible)?
Search: "[similar real object] assembly exploded view" for seam logic
Search: "[design language] panel line style ArtStation hard surface"
PANEL SIZE CALIBRATION:
Panels that are too small create visual noise. Panels that are too large
read as architectural rather than mechanical.
Rule of thumb:
- Largest panel: no more than 25–30% of the total asset's longest dimension
- Smallest visible panel: no smaller than 2× the bevel width
- Panel hierarchy: 3 sizes maximum — large structural / medium access / small detail
Reference: measure panel sizes on 5 existing assets in this design language.
Document the ratio between largest and smallest panel.4. Bevel calibration reference
The bevel is the single most important technical decision in hard surface. It determines how light reads across edges at any distance.
BEVEL FUNCTION:
A bevel catches specular highlights along edges.
Too narrow: bevel doesn't catch light — edge reads as infinitely sharp (unrealistic)
Too wide: bevel reads as soft, rounded — loses the "hard" in hard surface
Correct width: consistent, catches a highlight visible at the render distance
BEVEL WIDTH CALIBRATION:
Step 1: Find 3 reference assets in your design language at the same poly budget.
Step 2: Zoom in on edge regions in their wireframe screenshots.
Step 3: Count edge loops at the bevel — note whether it's 1, 2, or 3 loops.
Step 4: Compare to their final renders — what does that bevel width look like lit?
Your target bevel:
Width (edge loops): [1 = tight / 2 = medium / 3 = wide]
Angle: [45° is standard / sharper or softer changes the highlight shape]
Applied to: [all hard edges / primary edges only / no hard edges (SubD workflow)]
Bevel rule: establish the width before detailing begins. Write it down.
Apply it to every hard edge. Break the rule only at 2–3 focal points.
BEVEL REFERENCE QUERIES:
1. "[design language] hard surface bevel width reference ArtStation"
2. "[asset type] edge highlight render reference"
3. "[game title similar to target] weapon prop wireframe close-up"5. Fastener and hardware reference
Fasteners are where hard surface scale errors are most visible. A screw that is 30% too large destroys the mechanical read of an entire asset.
FASTENER INVENTORY:
List every fastener type visible on this asset:
[Fastener type 1]:
Real-world size: [diameter in mm at actual scale]
At texel density: [how large in pixels on the texture sheet]
Head type: [hex / Phillips / flathead / bolt / rivet / panel screw]
Placement logic: [where does this fastener logically appear on this object?]
Reference query: [specific macro photography query]
[Fastener type 2]: ...
FASTENER SCALE RULE:
M3 screw (3mm diameter): appropriate for small electronics and fine mechanisms
M6 screw (6mm): appropriate for medium panels and structural joins
M10+ bolt: structural, high-stress joins only
Rivet (3–6mm): sheet metal, aircraft construction, armor plating
Most game hard surface errors: fasteners are 2–3× too large.
Reference real hardware at the correct scale before modeling any fastener.
FASTENER PLACEMENT LOGIC:
Fasteners appear where stress is highest — corners, edge joins, panel corners.
They do not appear in the center of flat panels (no structural purpose there).
They appear at regular intervals along joins — not randomly spaced.
Search: "[object type] fastener placement engineering reference"
Search: "aircraft panel rivet spacing reference", "military hardware fastener reference"6. Greebling reference
Greebling is surface complexity added to suggest mechanical sophistication. Uncontrolled greebling reads as noise. Controlled greebling reads as purpose.
GREEBLING PHILOSOPHY FOR THIS ASSET:
□ No greebling — clean design, detail comes from form and material only
□ Minimal greebling — 1–2 zones of increased complexity, rest is clean
□ Moderate greebling — balanced complexity across the surface
□ Heavy greebling — dense complexity throughout (sci-fi capital ship level)
GREEBLING DENSITY MAP:
Map where complexity should sit using the 60/30/10 rule:
- 60% clean primary form: [which areas]
- 30% medium panel/mechanical detail: [which areas]
- 10% dense greebling: [which areas — typically cockpits, engine zones, focal points]
GREEBLING LOGIC:
Every piece of greebling should have an implied function:
- Vents and grilles: heat dissipation
- Recessed panels: access hatches, maintenance covers
- Protruding elements: sensors, ports, mounting points, emitters
- Cable runs: power and data routing between sub-systems
- Structural ribbing: load-bearing reinforcement under stress zones
Greebling without implied function reads as decoration.
Reference: "[spacecraft/vehicle type] technical reference book" for function logic
Reference: real military/industrial hardware for what mechanical complexity looks like
GREEBLING SCALE CALIBRATION:
Greebling must vary in scale to read correctly:
- Large greebling: sub-assembly level (visible at 5m+ camera distance)
- Medium greebling: panel and component level (visible at 1–3m)
- Fine greebling: surface detail (visible at 0–1m or in close-up screenshots)
All three levels must be present for greebling to read at multiple distances.7. Material reference
Hard surface materials have specific PBR requirements.
MATERIAL ZONES:
[Zone 1 — Primary body material]:
Material type: [e.g. painted steel / anodized aluminum / matte polymer]
Albedo range: [linear value range — calibrated to material class]
Roughness: [0.0–1.0 range — painted metal typically 0.55–0.70]
Metallic: [0 for painted / 1 for raw metal / note exceptions]
Wear behavior: [where paint chips / where metal is exposed / edge wear pattern]
Reference query: [specific macro photography query]
[Zone 2 — Secondary material]:
...
EDGE WEAR CALIBRATION:
Edge wear exposes the underlying material.
For each transition:
- What material is revealed? (raw metal / primer / different paint layer)
- What color is the exposed layer?
- How wide is the wear zone? (hairline scratch / broad chip / full exposure)
Search: "[primary material] edge wear chip reference close-up macro"
SURFACE IMPERFECTION LOGIC:
Hard surface imperfections follow physical law:
- Scratches: along the direction of contact (a knife mark runs with the blade)
- Impact dents: radial deformation around the impact point
- Corrosion: spreads from water contact zones, runs with gravity
- Wear: concentrates at contact surfaces (handles, edges, ground contact)
Never: random scratches that go in all directions with no physical motivation.8. Reference board layout
┌──────────────────────┬──────────────────────┐
│ DESIGN LANGUAGE │ SILHOUETTE & │
│ REFERENCE ASSETS │ PRIMARY FORM │
│ (10 images) │ (8 images) │
├──────────────────────┼──────────────────────┤
│ PANEL LINE SYSTEM │ BEVEL CALIBRATION │
│ (style + logic) │ (wireframe refs) │
│ (10 images) │ (6 images) │
├──────────────────────┼──────────────────────┤
│ FASTENER + HARDWARE │ GREEBLING REFERENCE │
│ (macro close-ups) │ (density examples) │
│ (12 images) │ (10 images) │
├──────────────────────┼──────────────────────┤
│ MATERIAL ZONES │ SCALE ANCHORS │
│ (per material) │ & ORTHO VIEWS │
│ (10 images) │ (8 images) │
└──────────────────────┴──────────────────────┘Target: 74 images. Design language and panel system are the priority zones these decisions cascade through every other reference category.
9. Pre-Modeling checklist
- Design language defined: 5 reference assets studied, rules extracted
- Primary form hierarchy documented, large/medium/small forms planned
- Silhouette reads clearly at 64×64px
- Panel line width decided and written down, one consistent rule
- Panel line logic established, assembly logic, not decoration
- Bevel width decided from reference, written as a rule
- Fastener inventory complete. Every type referenced at correct scale
- Greebling density map planned: 60/30/10 zones identified
- Material zones documented with PBR value ranges
- Edge wear logic documented, physical rules, not random
- Scale anchor confirmed, real-world dimensions in cm/m
- Function logic documented. I can explain what every major feature does
Output
Save as ref-hardsurface_[assetname]_v1.md in the project folder.
Run ref-brief alongside this skill for the full reference framework including orthographic views, lighting reference, and board layout. ref-hardsurface goes deeper on the hard-surface-specific layers that ref-brief covers at a general level.
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
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.

