What is a Design Skill?
A reusable instruction package that teaches AI agents how to handle design tasks. Open, vendor-agnostic, and portable across agent platforms.
A design skill is a reusable instruction package that teaches AI agents how to perform design tasks. Not a design tool. Not a component library. An instruction layer that sits between the agent and the design infrastructure it operates on.
Every design decision an AI agent makes follows a pattern. Choose a color. Apply a layout. Verify accessibility. These are repeatable workflows that can be encoded as portable instructions. A design skill packages these workflows into a format that agents can understand and execute across any compatible platform.
Unlike tools that require human operators, skills encode judgment into executable instructions. An agent that has access to the right skill knows not just what it is allowed to do, but how to do it correctly for the context it is operating within.
Three categories of design skills
Canvas Skills
Instructions for working within a specific design tool. Figma skills, Stitch skills. Powerful but locked to a single platform.
Portable Skills
Vendor-agnostic instructions that work across any agent platform. The open standard. No lock-in. Reusable anywhere.
Orchestration Skills
Instructions that chain multiple tools and services together. Fetch from Figma, validate through expression infrastructure, implement in code.
Skills vs infrastructure
Skills are the workflow layer. They tell agents how to sequence operations: step 1, step 2, step 3. They are procedural and actionable.
Expression infrastructure is the judgment layer. It tells agents what the right answer is. A skill might say “apply brand colors.” Infrastructure resolves which specific colors, in what relationships, with what contrast ratios, for what context. Both are needed. Neither works alone.
The agent with a skill but no infrastructure follows instructions blindly. The agent with infrastructure but no skill has judgment but no workflow. The complete system has both: orchestration from skills, wisdom from infrastructure.
Why portability matters
Canvas skills lock you into a specific tool. If your workflow changes, you rebuild. Portable skills are investment-safe. A skill written once can run on Cursor, Claude Code, Copilot, or any future AI builder.
Portability is the difference between a skill that works for today and a skill that works for the decade. It is the difference between building for a single vendor and building for a category.
Finding and sharing skills
skill.design maintains an open registry of design skills for AI agents. Browse, discover, and contribute skills that cover design system auditing, accessibility checking, token management, component generation, and cross-tool orchestration.
The registry is a commons. Contributions welcome. No vendor, no gatekeeping, no lock-in.