Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
A Claude Code plugin that teaches visual design — typography, color, composition, proportions, and visual hierarchy
| Date | Stars |
|---|---|
| 2026-07-31 | 270 |
| 2026-08-03 | 273 |
| 2026-08-06 | 273 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# design-for-ai Claude builds interfaces that look like Claude built them. Cyan-on-dark hero sections. Cards in a grid, all the same size. Glassmorphism on everything. Bounce easing. Inter at 16px. You've seen it. You've shipped it. Your users noticed. This plugin teaches Claude the design vocabulary it's missing. Not "make it pretty" — the actual frameworks working designers use: proportional systems that make element sizing feel right instead of arbitrary, type scales built from ratios instead of guessing, color palettes constructed from the color wheel instead of vibes, composition rules that create a visual anchor instead of uniform mush. Based on *Design for Hackers* by David Kadavy. Nine chapters of design theory, distilled into reference files that Claude loads on demand. ## A design-foundations system As of **v4.2.0**, this is a **design-foundations system** with one front door: a four-stage workflow that takes any design idea from brief to viewable artifact. Design knowledge lives as doctrine reference files the workflow loads deterministically — plus a `prototype` skill you can invoke directly to render a mock. ## The workflow ``` /design-for-ai:research → /design-for-ai:plan → /design-for-ai:mock → /design-for-ai:build ``` Start with an idea — vague is fine. The workflow takes it from brief to gated, committed design artifacts. | Command | What happens | |---------|-------------| | **research** | Facilitates the design brief: who's it for, what should it feel like, what's the job it does — and catches taste signals (fonts you lean toward, colors you'd never ship) as pins for the DNA deal. Saves a research doc; an optional verification pass (a grill, then a zero-context cold read) flips it draft → confirmed. | | **plan** | Turns the brief into a phased design plan: Discover (journey/IA/flows) → Design (DNA/tokens/system/words). Assigns doctrine per phase, sets design done-when criteria. | | **mock** | Renders a throwaway-fidelity prototype from the plan direction and runs a cheap cross-pillar validation on the real pixels. Gates on your sign-off before committing to the full build. | | **build** | Executes the approved plan phase by phase — each phase: `BUILD → REVIEW → commit` with dispatched subagents, design execution evidence (contrast/tokens/heuristic), and a final trust report. | The doctrine files are the workflow's knowledge library: each `mock` and `build` phase's dispatched agents resolve the applicable doctrine names and `Read()` every file before executing, so a cross-pillar review of real rendered pixels is a workflow stage, not a trigger guess. Review is dual-blind — an isolated cross-pillar critique and a deterministic AI-tell detector (`scripts/detect.mjs`) gather findings separately and synthesize after both finish. ### The doctrine domains Eight domains, loaded per-phase by the workflow (not directly invocable): | Domain | Concern | |--------|---------| | **usability** | whether users can *operate* it — Nielsen 10, UX laws (Fitts/Hick/Miller), affordances, cognitive load. The keystone the others cite | | **content-design** | the *words* are the interface — UX writing, microcopy, error/empty/button copy, voice & tone | | **data-viz** | encoding *data* — which chart, dashboards, data-ink/chartjunk, truthful encoding, chart accessibility | | **deceptive-patterns** | dark patterns, manipulative UX, an ethical design review, regulatory exposure (DSA/FTC). The twin of the AI-tells ban-list | | **behavioral** | *why* users act/return/convert — persuasion (Cialdini), the Fogg model, habit loops, emotional design. The honest mechanism | | **journey** | how a user *moves through time* — JTBD, journey maps, IA/sitemaps, user flows, page specs. Ships a JOURNEY.md | | **design-systems** | a look → a *machine* — design tokens, atomic components, governance. Extends DESIGN.md | | **ai-native** | agent/LLM-interface design — agent UX, generative UI, no-fixed-screen interfaces. Principle-derived; **no s
Excerpt of 7,134 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:d901a8c44e9cde98, llm:Description: 'A Claude Code plugin that teaches visual design — typography, color, composition, proportions, and visual hierarchy'. Language: JavaScript. No topics provided.
matched fp:d901a8c44e9cde98, llm:Description: 'A Claude Code plugin that teaches visual design — typography, color, composition, proportions, and visual hierarchy'. Language: JavaScript. No topics provided.
matched fp:d901a8c44e9cde98, llm:Description: 'A Claude Code plugin that teaches visual design — typography, color, composition, proportions, and visual hierarchy'. Language: JavaScript. No topics provided.