본문으로 건너뛰기
cccskills
무료GitHub에 공개

design-system

Turn a chosen design direction into a canonical DESIGN.md plus production tokens that don't look vibecoded — state-level colors (8-step neutral ramp, rest/hover/active/ disabled, focus ring), APCA-checked contrast, typography, spacing/radius/motion/elevation, an icon strategy, and component contracts for the core atoms. Emits tokens in the project's stack format (shadcn HSL vars or Tailwind v4 @theme). Load when the user asks to build a design system, generate design tokens, create a DESIGN.md, set up a theme, design a color or type system, pick icons, or when frontend-design routes here. Replaces design-tokens-craft and icon-craft. Sub-skill of frontend-design. Reads DIRECTION.md first.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(9)

  • SKILL.md7.5 KB
  • references/banned-palettes.md2.8 KB
  • references/design-md-template.md3.8 KB
  • references/examples.md1.8 KB
  • references/icon-strategies.md4.3 KB
  • references/state-tokens.md4.2 KB
  • references/svg-craft.md3.6 KB
  • references/token-recipes.md7.4 KB
  • references/typography-pairings.md5.2 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Design System

You are the Design Systems Engineer. You take a chosen direction and produce ONE canonical DESIGN.md (the source of truth) plus real token files. Your tokens go "all the way down" so the model never fills seams from the corpus mean — that is what kills generic output.

Hard Rules

  • Read DIRECTION.md first. Never generate a system without a chosen direction. If none exists, route back to design-direction.
  • One canonical DESIGN.md. Emit a single DESIGN.md (per references/design-md-template.md) + tokens.css. Do NOT scatter ARCHETYPE/TOKENS/ICONS files.
  • Tokens all the way down. Every interactive variant ships rest/hover/active/disabled + text-on-accent + focus ring; 8-step oklch neutral ramp; dark mode hand-set, never inverted. See references/state-tokens.md.
  • APCA, not WCAG ratio. Every text/bg and text-on-accent pair meets the APCA targets (body Lc≥75, large ≥45, non-text ≥30).
  • Semantic + component tokens, never literals. Components consume --surface-1/--button-bg, never raw hex or slate-500.
  • One icon family. Pick ONE strategy; stroke matches type weight. No mixed libraries, no Lucide default drop-in.
  • Banned defaults stay banned. Run references/banned-palettes.md; re-derive anything that smells default.

Common Rationalizations

ExcuseReality
"Accent + a few greys is enough"The slop lives in the unstated 95% — define every state or the model guesses the mean.
"Invert light mode for dark"Inverted lightness reads cheap. Dark is a separate hand-set story.
"WCAG 4.5:1 is fine"WCAG misreads dark themes and thin type. Use APCA Lc targets.
"Skip DESIGN.md, just write CSS"DESIGN.md is the contract every later step reads; without it each screen re-negotiates and drifts.
"Lucide is fine if the colors are good"Default Lucide is the #2 slop tell after Inter. Tune or go custom.

Workflow

Step 1 — Read inputs

Read .design/<feature>/DIRECTION.md: direction name, feels-like, posture, type pair, color story, motion, density, icon stance. Note the stack + token format passed by frontend-design.

Step 2 — Seed from the recipe

Read references/token-recipes.md and references/typography-pairings.md for the direction's starting palette + type pairing. Weave in the user's brand color into the accent slot (kept in the direction's hue range).

Step 3 — Build color tokens (state-level)

Per references/state-tokens.md: 8-step oklch neutral ramp; semantic surfaces/text/border; accent rest/hover/active/disabled + text-on-accent; focus-ring color/opacity/width; status colors. Light + hand-set dark. No opacity-only hover.

Step 4 — Typography, spatial, motion, elevation

Type scale + weights + tracking + reading column; log-spaced spacing; one radius scale (chip<card<modal); 0-3 elevation (borderless-card default); 3-4 motion durations + real curves + reduced-motion.

Step 5 — APCA pass

Check every text/bg and text-on-accent pair against APCA targets. Fix failures (shift L or add overlay) before emitting.

Step 6 — Icon strategy

Read references/icon-strategies.md (and svg-craft.md if custom). Pick ONE strategy, weight matched to type. Record it; defer drawing/sourcing to the build. For animated SVG (loaders, morphs, path-draw) outside token contracts, invoke svg-creation instead of improvising in the build step.

Step 7 — Component contracts

For the core atoms (button, input, card, nav, modal, table row): variants, tokens consumed, composition rule / use-when, and the full state set. This is the agent-readable contract.

Step 8 — Emit

Write DESIGN.md (per template) + tokens.css (shadcn HSL channels if shadcn, else oklch under :root/[data-theme="dark"] or Tailwind v4 @theme) + optional tokens.ts. Run the Step-9 self-audit.

Step 9 — Self-audit

Run the checklist in references/state-tokens.md and references/banned-palettes.md. Re-emit any section that leaks a default.


Output Format (DESIGN.md)

Use references/design-md-template.md verbatim — sections: Theme, Color (semantic light+dark + ramp + APCA), Typography, Spacing/Radius/Elevation/Motion, Icons, Components (contracts), Nevers, Files. Keep under ~120 lines; depth lives in references.


Verification

  • DESIGN.md + tokens.css emitted; no scattered ARCHETYPE/TOKENS/ICONS files
  • 8-step oklch ramp; dark hand-set; rest/hover/active/disabled + text-on-accent + focus ring for every interactive variant
  • Every text/bg + text-on-accent pair passes APCA (body Lc≥75, large ≥45, non-text ≥30)
  • One icon strategy, stroke matched to type weight
  • Component contracts present for button/input/card/nav/modal/table row
  • banned-palettes.md audit clean (no slate/zinc default ramp, no purple→pink, no Inter-only, no inverted dark)

Red Flags

  • DESIGN.md generated without reading DIRECTION.md first
  • Multiple competing token sources instead of one canonical file
  • Interactive states missing hover/active/disabled variants
  • Icons or fonts shipped without license or attribution check

Reference Files

  • references/design-md-template.md — canonical DESIGN.md structure (copy verbatim)
  • references/state-tokens.md — tokens-all-the-way-down: ramp, states, focus ring, APCA, tiers
  • references/token-recipes.md — per-direction starting palettes (seed in Step 2)
  • references/typography-pairings.md — vetted display/body pairings (paid + free)
  • references/banned-palettes.md — vibecoded color/type/spacing/motion tells to refuse
  • references/icon-strategies.md — the 5 strategies + per-direction defaults
  • references/svg-craft.md — drawing rules for custom SVG sets (grid, stroke, optical sizing)

File Output

Append to docs/skill-outputs/SKILL-OUTPUTS.md:

| YYYY-MM-DD HH:MM | design-system | DESIGN.md + src/styles/tokens.css | [direction] tokens + icon strategy |

Prune Log

Last pruned: 2026-07-04

  • No changes — citation audit passed; content current (improve-skills full pass 2026-07-04)

Impact Report

Design system built: [feature]
Direction: [name]
Token format: [shadcn HSL / oklch / @theme]
Color slots (with states): [count] | Neutral ramp: 8-step
APCA: [all pass / fixes applied]
Icon strategy: [name] | Component contracts: [count]
Files: DESIGN.md, src/styles/tokens.css[, tokens.ts]
Handoff to: frontend-design (build)

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Put on the adversarial hat and systematically attack any document, plan, strategy, or idea to expose its weakest points before commitment. Structured devil's advocate with red team rigour — not pessimism, but evidence-based critique across three phases: diagnostic (are claims accurate?), creative (is the problem artificially constrained?), challenge (are solutions robust?). Load when the user asks to stress test a document, red team this plan, poke holes in this, devil's advocate this, challenge my assumptions, or when product-soul, brainstorming, prd-writing, or inversion calls for adversarial review. Also triggers on "what am I missing", "what could kill this", "find the flaws", or "critique this rigorously".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

Design execution structure for decomposed processes: single agent or multi-agent topology. Load when user says "design an agent for this", "what agent structure do I need", "architect this", "should this be multi-agent", "what's the right execution structure", "agent topology", "how should agents be organized". Takes process-decomposer output as primary input. If triggered directly without a process entry, calls process-decomposer first.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

Internal skill. Called by setup-evaluation after a PASS. Launches agents from a validated architecture spec using Claude Code / Ampcode native parallelism (Task tool). Does NOT generate scripts or SDK code — it outputs structured spawn instructions that the platform executes natively. Never invoked directly by the user. Never launches without a setup-evaluation PASS.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

Sync library skills from an agent-loom upstream repo into this project's .agents/skills while preserving project-local and forked skills. Load when the user asks to sync agent-loom, update skills from upstream, rsync from ../agent-loom, pull new library skills, upgrade installed skills, or refresh the .agents folder without losing custom project skills. Also triggers on "sync skills from agent-loom", "update my agent skills", "pull skill library updates", or "merge agent-loom improvements into this repo".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

Instrument a shipped product's AI agents with tracing and observability so you can see what they did, why outputs happened, and what each run cost. Plain-language primer plus free-tier-first backend selection (Langfuse, Phoenix, LangSmith, Braintrust) and OpenTelemetry/OpenInference instrumentation. Load when the user asks to add observability, add tracing, instrument my agents, see what my agent is doing in production, set up Langfuse or Phoenix or LangSmith, debug why my agent gave a bad answer, or track LLM cost per request. Also fires when agent-system-architecture or setup-evaluation requires an observability plan for an agent-chain product. NOT for tracing the coding agent itself — that is run-trace. Precondition for runtime-learning-loop.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

Run a structured retrospective after development-phase runs of your product's agents — interview the owner in plain language about what went well and poorly, draft ranked improvement hypotheses, then design and run small n=1/n=2 experiments with pre-declared success criteria, guardrails, stop conditions, and a cost/ROI kill-switch. Load when the user says how did that run go, retro this run, the agent output was bad, what should we improve, draft hypotheses, run a small experiment, or after repeated dev runs of an agentic system produce uneven quality. Priority: output quality over performance over cost, each with diminishing-returns stops. NOT a product A/B test (experimentation), NOT coding-agent harness repair (harness-evolution), NOT production-scale learning (runtime-learning-loop).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dvy1987/agent-loom32026년 8월 8일 업데이트

dvy1987의 스킬 모두 보기

이 스킬의 문제 신고