Standalone assets that belong to the brand, authored as SVG in code. The skill reads your brand truth first and uses only recorded values; restraint is the default register, and the typographic answer comes before any drawn element.
Say this, get this
| You say | You get |
|---|---|
| “Make an SVG header image for this post” | One atmospheric asset from your recorded brand values |
| “Make the repo’s social image” | SVG as the source, PNG at 1280×640 for GitHub and 1200×630 for OG |
| “Logo concepts” | Two or three genuinely different candidates, each with its method named |
| “An on-brand image of the new feature” | A concept image; an interactive mock is pe-design |
Modes
This skill has no modes: one workflow, run in phases.
Setup
required |
brand truth: DESIGN.md, a style guide, or a brand/ folder; with none of these, the skill offers pe-design’s understand mode or proceeds only on values you supply, labeled provisional |
optional |
an SVG-to-PNG renderer, for OG and social images |
convention |
SVG is the source of truth; PNG only where a destination requires it |
What it produces
SVG files with a real viewBox and named groups, plus PNG exports where a destination
requires raster, and a report: the tokens and motifs used, the method chosen, placement,
sizes, and anything provisional that should graduate into DESIGN.md. Every asset is
critiqued against the anti-generic tells and redone before it ships.
Handoffs
No brand truth: pe-design, understand mode. The asset becomes a UI component: pe-build. Judging existing assets: pe-review. Animating an asset: pe-build, motion mode.
Provenance
Carries 1 files from Leon (1). Every vendored file is hash-locked to a pinned upstream commit.
Apache-2.0 · skills/pe-brand-assets