Skip to content

vaoferi/vaoferi-design-skill

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

57 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Vaoferi Design Skill

Design with structure, not decoration. Use grids before colors, alignment before effects, components before custom code, and trace existing CSS before writing new CSS.

Now the skill also starts design-system work with a portable DESIGN.md contract, so tools like Google Stitch, Open Design, Codex, Claude Code, Cursor, and similar coding agents can read the same source of truth instead of guessing the UI.

Main rules

  • Start from goal and hierarchy.
  • Create or reuse DESIGN.md before designing multi-screen UI.
  • Start and fill the primitive library before generating page-specific elements.
  • Build the grid first using Golden Canon-inspired structure.
  • Reuse components before creating new ones.
  • Match the existing UI first on operational screens and admin forms.
  • Trace existing CSS before adding a new selector or override.
  • Treat !important as an emergency escape hatch, not a normal fix.
  • Use design tokens before ad-hoc values.
  • Ask before introducing new components, colors, or spacing scales.
  • Validate responsive behavior across breakpoints.
  • Finish only after visual QA.
  • Stop on conflicts and show the Conflict Form before proceeding.

Teaser

This skill turns 1: 618, 1fr 1fr, and Fibonacci spacing into a workflow that produces layouts which:

  • behave the same on mobile and desktop;
  • reuse existing UI and existing CSS instead of inventing fresh styles;
  • keep spacing and typography consistent through tokens.

Quick start (recommended flow)

goal
  -> DESIGN.md
  -> primitive library
  -> hierarchy
  -> Golden Canon-inspired grid
  -> alignment lines
  -> component resolution
  -> existing components/tokens
  -> proposal if something new is required
  -> final styling
  -> visual QA

If sources or libraries contradict each other, stop and use the Conflict Form from SKILL.md before continuing. If a new component, token, or layout pattern is required, use the Approval Flow from SKILL.md.

Use

  1. Understand the goal and user action.
  2. Create or find DESIGN.md.
  3. Start the primitive library: Button, Card, Input, Text, Grid, Container, states.
  4. Build the grid before placing elements.
  5. Draw alignment lines.
  6. Resolve components before creating new ones.
  7. Trace existing CSS before adding new styles.
  8. Ask approval before adding new styles or CSS escape hatches.
  9. Check responsive behavior.
  10. Protect photos and logos.
  11. Use design tokens only.
  12. Verify in browser or preview before marking Done.

What this skill does

  • prevents random one-off CSS and component sprawl in operational/admin UI;
  • creates a portable DESIGN.md contract that Stitch, Open Design, Codex, Claude Code, and Cursor can all read;
  • enforces token discipline so colors, spacing, radius, and typography stay consistent;
  • applies Golden Canon-inspired structure so layouts have predictable rhythm;
  • keeps flat, traceable class naming and clear CSS ownership;
  • tracks real usage and periodically uses SkillOpt-style review to improve the skill without auto-replacing SKILL.md;
  • requires visual QA with explicit Pass / Fail / Partial evidence instead of "looks fine";

What this skill does NOT do

  • this skill will not fix a broken product by adding decoration over a bad structure;
  • it will not approve new colors, components, or spacing without an explicit proposal;
  • it will not skip DESIGN.md because "we can just style the page directly";
  • it will not treat overflow-x: hidden or fixed heights as acceptable layout fixes;
  • it will not accept best_skill.md from SkillOpt without human diff review and validation evidence;
  • it will not claim a design is Done without visual evidence.

What this skill emphasizes

  • DESIGN.md is the shared contract for Stitch/Open Design-style agents.
  • Tokens must be mirrored to code-friendly files like tokens.css / tokens.json when needed.
  • Primitive libraries prevent one-off buttons, cards, inputs, and badges on every page.
  • SkillOpt is used as a periodic improvement loop: real traces, scored examples, validation gate, reviewed best_skill.md.
  • Golden Canon is a macro guide, not a rigid pixel rule.
  • grid, subgrid, fr, minmax(), auto-fit, and clamp() are the main primitives.
  • grid handles page structure; flex handles the interior of components.
  • gap is preferred over margin hacks.
  • CSS guardrails: trace before patching, low specificity, no !important by default.
  • Traceable class naming and @container help keep component ownership and local responsiveness clear.
  • Accessibility and responsive scaling are first-class requirements.
  • Mobile-first defaults start from the narrowest supported width, keep touch targets usable, and prefer rem, box-sizing, and intentional local scroll over cramped desktop fallbacks.
  • Visual polish comes from rhythm, whitespace, typography, and restrained motion, not extra decoration.
  • The browser should do the math whenever possible.

Files

  • SKILL.md — main skill entrypoint
  • AGENTS.md — rules for working in this repository
  • README.md — project overview
  • rubric.md — Done checklist for design work
  • docs/history/project_log.md — log of notable skill changes and decisions
  • examples/good-answer.md — приклад правильної відповіді за скілом
  • examples/bad-answer.md — приклад типових помилок

About

AI design skill for building responsive layouts with Golden Canon-inspired grids, reusable design components, design tokens, and strict visual QA.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors