From ih-designer
Help a project discover its own design aesthetic and build its own design system from scratch — when it has none, or only scattered fragments. Use when starting design work on a project without a coherent system, or when the user wants to establish one. Analyzes the target project, asks sharp questions to shape a direction, recommends options with rationale, then builds the tokens, type, color, spacing, and conventions INTO the target project's repository. There is no house baseline to impose — every value is chosen for this project. Confirm maturity with detect-design-system first.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ih-designer:build-design-systemThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
When a project has no design system of its own (or only scattered, inconsistent fragments), designer's job is to help *that project* find its aesthetic and build its own system — not to hand it a generic one. **Designer has no baseline design system to impose.** Every color, type choice, and spacing value is chosen for this project, and everything you produce is integrated into the **target pro...
When a project has no design system of its own (or only scattered, inconsistent fragments), designer's job is to help that project find its aesthetic and build its own system — not to hand it a generic one. Designer has no baseline design system to impose. Every color, type choice, and spacing value is chosen for this project, and everything you produce is integrated into the target project's repository, not designer's.
Run detect-design-system first. Only build when maturity is None or Partial (for Partial, consolidate toward the strongest existing direction rather than starting over). If a coherent system already exists, use design-audit instead — don't rebuild what's working.
Understand what you're designing for before proposing anything:
brand-kit), competitor/reference products.A design system is the project's, so shape it with the user. Ask a small number of high-leverage questions, never a blank "what do you want" and never a long dump:
Then offer 2–3 concrete aesthetic directions, described specifically (not "modern" or "clean"), each with a rationale tied to the goals. Show, don't abstract — name the type feeling, the color temperature, the density, the motion stance. Let the user react and choose; converge on one.
Before generating, declare it in one line:
"Building this as: a [kind] for [audience], with a [vibe] language — [type feeling], [color direction], [density], [motion stance]. Goal: [outcome]."
Confirm this is right, then build.
Produce a real, coherent, minimal-but-complete system and write it in the target project's native stack (Tailwind config, CSS custom properties, a tokens file, MUI/Chakra theme — whatever the project uses):
DESIGN.md (or the project's convention) in the target repo capturing the system and the why, so the team and future sessions can extend it consistently.Keep it a coherent small system, not a 40-page guide. Ship tokens + a few primitives + a page of docs.
brand-kit — capture or define the project's identity; the system encodes it as tokens.ux-principles — the quality floor every system must clear.impeccable, taste-skill, emil-design-skills — reach into these for craft, taste direction, and motion detail while building. They raise quality; they don't decide the project's identity — the project does.Never impose a generic or reused aesthetic. Two projects built with this skill should end up looking like themselves, not like each other. If you catch yourself reaching for a default palette, type pairing, or the last project's system, stop and go back to the read.
npx claudepluginhub weberswords/designer --plugin ih-designerGuides completion of development work by verifying tests, detecting environment, and presenting structured options for merge, PR, or cleanup.
Guides creation and editing of skills using test-driven development with pressure scenarios and subagents to verify agent compliance.
Dispatches multiple subagents concurrently for independent tasks without shared state. Use when facing 2+ unrelated failures or subsystems that can be investigated in parallel.