anatomy-asset
Make crafted, highly detailed isometric SVG assets for landing pages, product sites, launch posts, social cards and decks — a hero object, a set of spot illustrations, a feature-tile scene — drawn as real, well-made physical things with many small precise solid parts, four-tone shading, correct painter's order and calm CSS-only ambient motion, mostly monochrome with brand accents. Use this when someone wants a nice-looking illustration, hero image, product visual, spot illustration, icon set, isometric artwork, "Linear-style" or "premium" drawing, or a visual for a section of a site, where the goal is a striking, on-brand picture rather than a diagram that explains how something works. Ships a standalone .svg that works as <img>, inline or in design tools, plus a preview page; can add WebGL shaders or the 3D mode when asked. For interactive figures that explain a technical concept with a live readout, use the anatomy skill instead.
Install
npx skills add https://github.com/Teddarific/skills --skill anatomy-assetComposition
Where the art sits on its canvas, and how it holds a page. An Anatomy figure is centred in a card; an asset is placed on a canvas beside a headline, inside a tile, or in a row of siblings.
Contents
- Canvases
- Regions and safe zones
- Placing the object
- Cameras for assets
- Sets that match
- Line weight and small sizes
- Edges, bleed and the ground
1. Canvases
Start from the placement, not a size. Defaults:
Ask for the real size if it matters, and keep the viewBox at about 1 unit per CSS px at the size it's shown, so the line weight means what you set.
2. Regions and safe zones
- Fit the art into its region with
fitInto(points, region, { pad, azimuth })(kit/asset.mjs). Pass every extreme point:boxCornersof the base and the tallest tip, and the top of anything lifted. - Give
previewHtmlthesaferegion. The preview logs a console error when the art's bounding box enters it, socapture.mjsshowsconsole: errorsinstead ofclean. Open the page with?safeto draw the zone. - Halos and blurs count as art. If the halo grazes the safe zone, shrink the region, not the halo.
- Leave the safe zone empty: no floor grid, no texture, no shadow running under the text.
3. Placing the object
- Fill the region, not the canvas. In its region the object takes 80–90% of the width (the pad does the rest); on the canvas a hero object is typically 45–60% of the width.
- Lead toward the text. A hero on the right reads best when its long face or its moment faces left, toward the headline. Choose the azimuth for that (section 4).
- Strong points. Put the moment (
concepts.md, section 6) about a third of the way in from the region's edges, not dead centre and not at the rim. - Air. At least 24 units of empty canvas between the art and any edge it doesn't bleed off.
- Lifted parts add height: include their top in the fit points, and leave room above for the float stroke.
4. Cameras for assets
The kit camera is a fixed iso-like view: azimuth turns the object on the floor, elevation tilts the view.
- Azimuth. 45° shows two faces equally. 55–65° shows the long +y face (the front of a keyboard, a counter, a shelf) and turns its length across the screen; 25–35° shows the short +x end. Try two or three and keep the one with the most interesting face toward the viewer and toward the text.
- Elevation. 30° by default. Raise it (40–50°) to look into trays, drawers, open boxes and dioramas; never below 25°, where tops vanish and faces stack.
- One camera per set. Spots in a row share azimuth, elevation and world scale.
5. Sets that match
A row of three spots, or a hero plus feature tiles, must look like one family:
- One world scale (
P.scalewithin about 10% across the set). Don't fit each object to its box: fit the biggest, then build the others at the same scale and centre them. - One camera, one ground colour, one accent family; each spot has the accent on its own moment.
- One part vocabulary: the same screw dots, bevel sizes, foot style and label plates.
- Similar visual weight: count the solids, and add parts to the sparse one, never remove detail from the dense one.
- Build them from one script with a shared module for the camera and the shared parts, as Anatomy's examples share
iso-kit.
6. Line weight and small sizes
The kit draws strokes with vector-effect: non-scaling-stroke, so --anatomy-weight is in screen px whatever the scale.
At small sizes cut part count density rather than letting detail turn to mush: bigger screws (0.6), fewer ticks, no knurls under 4 units radius. Capture at the shipped size and at 390 px wide for phones, and look.
7. Edges, bleed and the ground
- Floating object (default). The object stands on its halo on the page colour. No floor, no frame.
- Ground plane. A tray, mat, slab, tabletop or cutting board under the object gives a scene its own floor. It is a real object with thickness and edges, never an infinite grid.
- Bleed. An object may run off the canvas edge (a long bench, a conveyor), but only along a clean straight edge of the object, and never off the edge next to the safe zone.
- Background.
assetSvg({ background: false })is the default: transparent between parts, so the page shows through. Usebackground: trueonly when the asset must carry its own rectangle (an OG card, a slide).