Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create and render OpenSCAD 3D models. Generate preview images from multiple angles, extract customizable parameters, validate syntax, and export STL files for 3D printing platforms like MakerWorld.
.claude/skills/dicklesworthstone-openscad/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 102% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 70% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 23% | 0% |
Create, validate, and export OpenSCAD 3D models. Supports parameter customization, visual preview from multiple angles, and STL export for 3D printing platforms like MakerWorld.
OpenSCAD must be installed. Install via Homebrew:
bashbrew install openscad
This skill provides several tools in the tools/ directory:
bash# Generate a single preview image ./tools/preview.sh model.scad output.png [--camera=x,y,z,tx,ty,tz,dist] [--size=800x600] # Generate multi-angle preview (front, back, left, right, top, iso) ./tools/multi-preview.sh model.scad output_dir/
bash# Export to STL for 3D printing ./tools/export-stl.sh model.scad output.stl [-D 'param=value']
bash# Extract customizable parameters from an OpenSCAD file ./tools/extract-params.sh model.scad
bash# Check for syntax errors and warnings ./tools/validate.sh model.scad
Always validate your OpenSCAD models visually after creating or modifying them.
After writing or editing any OpenSCAD file:
multi-preview.shread toolThis catches issues that syntax validation alone cannot detect:
Never deliver an OpenSCAD model without visually confirming it looks correct from multiple angles.
Write OpenSCAD code with customizable parameters at the top:
openscad// Customizable parameters wall_thickness = 2; // [1:0.5:5] Wall thickness in mm width = 50; // [20:100] Width in mm height = 30; // [10:80] Height in mm rounded = true; // Add rounded corners // Model code below module main_shape() { if (rounded) { minkowski() { cube([width - 4, width - 4, height - 2]); sphere(r = 2); } } else { cube([width, width, height]); } } difference() { main_shape(); translate([wall_thickness, wall_thickness, wall_thickness]) scale([1 - 2*wall_thickness/width, 1 - 2*wall_thickness/width, 1]) main_shape(); }
Parameter comment format:
// [min:max] - numeric range// [min:step:max] - numeric range with step// [opt1, opt2, opt3] - dropdown options// Description text - plain descriptionbash./tools/validate.sh model.scad
Generate preview images to visually validate the model:
bash./tools/multi-preview.sh model.scad ./previews/
This creates PNG images from multiple angles. Use the read tool to view them.
bash./tools/export-stl.sh model.scad output.stl # With custom parameters: ./tools/export-stl.sh model.scad output.stl -D 'width=60' -D 'height=40'
Common camera angles for previews:
--camera=0,0,0,45,0,45,200--camera=0,0,0,90,0,0,200--camera=0,0,0,0,0,0,200--camera=0,0,0,90,0,90,200Format: x,y,z,rotx,roty,rotz,distance
For MakerWorld, you typically need:
export-stl.shConsider creating a model.json with metadata:
json{ "name": "Model Name", "description": "Description for MakerWorld", "parameters": [...], "tags": ["functional", "container", "organizer"] }
bash# 1. Create the model (write .scad file) # 2. Validate syntax ./tools/validate.sh box.scad # 3. Generate multi-angle previews ./tools/multi-preview.sh box.scad ./previews/ # 4. IMPORTANT: View and validate ALL preview images # Use the read tool on each PNG file to visually inspect: # - previews/box_front.png # - previews/box_back.png # - previews/box_left.png # - previews/box_right.png # - previews/box_top.png # - previews/box_iso.png # Look for geometry issues, misalignments, or unexpected results. # If anything looks wrong, go back to step 1 and fix it! # 5. Extract and review parameters ./tools/extract-params.sh box.scad # 6. Export STL with default parameters ./tools/export-stl.sh box.scad box.stl # 7. Export STL with custom parameters ./tools/export-stl.sh box.scad box_large.stl -D 'width=80' -D 'height=60'
Remember: Never skip the visual validation step. Many issues (wrong dimensions, boolean operation errors, inverted geometry) are only visible when you actually look at the rendered model.
openscadcube([x, y, z]); sphere(r = radius); cylinder(h = height, r = radius); cylinder(h = height, r1 = bottom_r, r2 = top_r); // cone
openscadtranslate([x, y, z]) object(); rotate([rx, ry, rz]) object(); scale([sx, sy, sz]) object(); mirror([x, y, z]) object();
openscadunion() { a(); b(); } // combine difference() { a(); b(); } // subtract b from a intersection() { a(); b(); } // overlap only
openscadlinear_extrude(height) 2d_shape(); rotate_extrude() 2d_shape(); hull() { objects(); } // convex hull minkowski() { a(); b(); } // minkowski sum (rounding)
openscadcircle(r = radius); square([x, y]); polygon(points = [[x1,y1], [x2,y2], ...]); text("string", size = 10);
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | fail→pass | 14,758 | 7,741 | -48% | 1 | 1 | 0% | 2,428 | 3,271 | +35% | 0 | 0 | — |
case-17 | fail→pass | 6,486 | 2,942 | -55% | 1 | 1 | 0% | 1,129 | 2,277 | +102% | 0 | 0 | — |
case-01 | fail→fail | 71,602 | 11,392 | -84% | 1 | 1 | 0% | 5,242 | 2,110 | -60% | 0 | 0 | — |
case-02 | fail→fail | 28,426 | 5,918 | -79% | 1 | 1 | 0% | 6,257 | 2,068 | -67% | 0 | 0 | — |
case-03 | fail→fail | 23,311 | 4,082 | -82% | 1 | 1 | 0% | 4,491 | 2,051 | -54% | 0 | 0 | — |
case-04 | fail→pass | 11,156 | 11,421 | +2% | 1 | 1 | 0% | 2,131 | 3,393 | +59% | 0 | 0 | — |
case-05 | pass→pass | 11,016 | 11,396 | +3% | 1 | 1 | 0% | 1,822 | 3,941 | +116% | 0 | 0 | — |
case-06 | fail→pass | 8,528 | 5,698 | -33% | 1 | 1 | 0% | 1,707 | 2,898 | +70% | 0 | 0 | — |
case-07 | pass→pass | 9,562 | 4,579 | -52% | 1 | 1 | 0% | 1,488 | 2,474 | +66% | 0 | 0 | — |
case-08 | fail→pass | 9,814 | 3,541 | -64% | 1 | 1 | 0% | 1,895 | 2,322 | +23% | 0 | 0 | — |
case-09 | pass→pass | 13,657 | 2,439 | -82% | 1 | 1 | 0% | 2,137 | 2,251 | +5% | 0 | 0 | — |
case-10 | fail→pass | 10,474 | 2,099 | -80% | 1 | 1 | 0% | 1,611 | 2,083 | +29% | 0 | 0 | — |
case-11 | fail→pass | 9,777 | 2,641 | -73% | 1 | 1 | 0% | 1,604 | 2,230 | +39% | 0 | 0 | — |
case-12 | fail→pass | 12,007 | 2,673 | -78% | 1 | 1 | 0% | 1,802 | 2,291 | +27% | 0 | 0 | — |
case-13 | fail→fail | 9,404 | 1,668 | -82% | 1 | 1 | 0% | 1,465 | 2,083 | +42% | 0 | 0 | — |
case-14 | fail→fail | 12,796 | 19,679 | +54% | 1 | 1 | 0% | 2,327 | 2,049 | -12% | 0 | 0 | — |
case-15 | pass→pass | 27,031 | 14,621 | -46% | 1 | 1 | 0% | 4,376 | 4,411 | +1% | 0 | 0 | — |
case-16 | fail→pass | 3,794 | 1,192 | -69% | 1 | 1 | 0% | 648 | 1,965 | +203% | 0 | 0 | — |
case-19 | pass→pass | 4,196 | 3,586 | -15% | 1 | 1 | 0% | 607 | 2,339 | +285% | 0 | 0 | — |
case-20 | pass→pass | 8,131 | 9,466 | +16% | 1 | 1 | 0% | 1,330 | 3,407 | +156% | 0 | 0 | — |
case-21 | pass→pass | 14,130 | 17,259 | +22% | 1 | 1 | 0% | 2,728 | 4,333 | +59% | 0 | 0 | — |
case-22 | pass→pass | 26,062 | 9,661 | -63% | 1 | 1 | 0% | 2,087 | 3,697 | +77% | 0 | 0 | — |
case-23 | pass→pass | 33,775 | 19,488 | -42% | 1 | 1 | 0% | 3,107 | 5,561 | +79% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 19 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +39 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.