A curated collection of reusable skills for AI design agents. Each skill is a structured markdown file that describes a discrete design operation — from laying out grids to generating accessible color palettes.
Skills are loaded by design agents at runtime. An agent reads the skill file, understands the operation, its inputs and constraints, and executes it against the current canvas or design file.
For platform customers: fork this repo, tweak existing skills or add your own, and connect it to your workspace. Your agents will pick up the changes automatically.
skills/
├── layout/ Spatial arrangement, grids, auto-layout, spacing
├── styling/ Color, fills, shadows, effects, theming
├── typography/ Type scales, text styles, font pairing
├── components/ Component creation, variants, slots, overrides
├── accessibility/ Contrast checks, ARIA annotations, readability
├── export/ Design-to-code, asset export, handoff
└── prototyping/ Interactions, transitions, micro-animations
Every skill is a single .md file with these sections:
| Section | Purpose |
|---|---|
| Title & description | What the skill does in plain language |
| Parameters | Inputs the agent needs (with types and defaults) |
| Behavior | Step-by-step instructions the agent follows |
| Constraints | Guardrails, edge cases, things to avoid |
| Examples | Concrete before/after or input/output samples |
- Pick the right category folder (or create a new one).
- Copy an existing skill as a starting template.
- Be explicit about defaults — agents should never have to guess.
- Include at least one example with concrete values.
- Keep each skill focused on a single operation.
MIT