本文へ移動
cccskills
無料GitHub で公開

committed-aesthetic

How to write — and how to use — a skill that IS one aesthetic rather than a catalogue of them. A catalogue lets an agent pick, and it picks the modal option; a committed aesthetic makes it execute one thing precisely, against rules you can check. Use when a design keeps coming out competent and forgettable, when starting a new brand surface with no reference, or when authoring a house style you want executed the same way twice.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.6 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

A committed aesthetic beats a catalogue

The problem this exists for

ui-ux-pro-max ships 84 styles, 161 palettes and 57 font pairings, and design-advisor is wired to all of it. The catalogue is not the bottleneck.

The bottleneck is that a catalogue lets a model pick, and an unconstrained pick is the modal one. Asked for a landing page with no brand, any model reaches for the same safe centre: a soft gradient, a large sans heading, three rounded cards with a left accent border, generous neutral grey. Every input in the training distribution votes for it. Nothing in a catalogue entry saying "Swiss minimalism — clean, grid-based, generous whitespace" is specific enough to override that, because that sentence describes the safe centre too.

Compare a rule from a committed aesthetic:

every dimension is calc(fraction * 100vw) against a baseline width, so one page at 1024px reads as a perfect scaled-down of the same page at 1920px

You can check that. A design either obeys it or does not. That is the whole difference: a catalogue entry is a label, a committed aesthetic is a constraint.

The form

A committed-aesthetic skill has five parts. The first is the one that does the work.

1. Transferable DNA — numbered, specific, falsifiable

Between 8 and 14 rules. Each must be checkable by someone who was not in the room. Test every rule against this: could two designers disagree about whether a page obeys it? If yes, it is a vibe, rewrite it.

Vibe (useless)Constraint (usable)
"thin, elegant type""weights 100 or 300 only, never 400+"
"lots of whitespace""~60% of any above-the-fold area is empty"
"restrained motion""no shadows, no gradients, no bounce; one shadow exists in the system, on toasts"
"technical feel""product codes use en-dash U+2013, never hyphen"
"dark and dense""five surface steps, each ~5% apart; depth is a surface step, never a shadow"

2. What this does NOT do — the IP boundary, as a table

An aesthetic is inspired-by, not cloned. Name the specific assets that belong to whoever you are borrowing from — wordmark, icon set, information architecture, product names, copy — and for each give the reason and what to do instead. This is not a legal disclaimer at the bottom; it is a working table an agent reads while building, and it is the difference between an homage and a trademark problem. Where the source is a real company, default to inspired-but-distinct and ask before taking the clone path.

3. A component library that runs

One HTML file the agent can open, with every component rendered and its own CSS inline, copy-pasteable. A token table in prose is read and paraphrased; a rendered gallery is copied exactly. This is why the form works at all.

4. Templates for the signature patterns

The two or three moves nobody would reconstruct from the DNA list alone — the navigation shape, the section header, the hero. With swap these for your own brand marked in comments.

5. Reference captures, marked "study, do not pixel-copy"

Screenshots are for understanding the patterns. Building FROM a screenshot rather than from the rules produces a bad tracing of one page instead of a design in the family.

The banned list, and why we needed one

ui-ux-pro-max's anti-pattern tables are about accessibility, touch targets and performance. They are good and they are not about taste. Nothing in this repository told an agent what cliché looks like, so the only defence against slop was one sentence in design-advisor: "no generic clean modern minimal filler".

Banned by default on any surface without an existing system that calls for them:

  • A gradient as the page ground. Especially purple-to-blue, especially at 135°.
  • The card triplet: three equal cards, rounded corners, a coloured left border, an icon in a tinted circle.
  • Inter, Roboto, Arial, Open Sans, Fraunces, Poppins as the display face. Not because they are bad — because they are the default vote.
  • Emoji as iconography. Draw the icon. (Emoji are fine if the brand uses them.)
  • The hero that describes a category rather than taking a position. "The modern platform for X" is the sentence a model writes when it has not been told what to believe.
  • Glassmorphism, neumorphism, and a mesh-gradient blob unless the brief names one of them.
  • Symmetric three-column feature grids as the default second section.

A banned list is not a style. It removes the centre so a choice has to be made.

Using one

  1. If a system already governs, this skill does not apply — match what exists.
  2. If nothing governs and you can ask: offer 2–4 genuinely different directions, each an axis you can name, before building. Not five shades of one aesthetic — that is not a choice.
  3. If nothing governs and you cannot ask: commit to ONE, state the assumption in a line at handover, and put 1–2 low-fi alternates beside the deliverable, never instead of it.
  4. Once a direction is chosen, load its committed-aesthetic skill and execute the DNA rules literally. Cite the rule number for each choice, the way design-advisor already cites a ui-ux-pro-max rule.

Instances in this repository

Related: skills/ui-ux-pro-max/SKILL.md is the catalogue this stands beside, not a replacement for it — the catalogue is where you look when you need to KNOW something; this is what you load once you have DECIDED something.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

great_cto's own committed aesthetic — the instrument panel. Dark five-step surface ladder, exactly one accent, two faces divided by MEANING (Geist speaks, Geist Mono is machine-truth), tabular numerals, and a dash that is not a nought. Twelve checkable rules measured from packages/board and greatcto.systems, not designed for this file. Use when building or extending any great_cto surface — the board, the site, a report, a share page.

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

Catalogue of known SDLC anti-patterns that great_cto agents must actively reject when reviewing architecture, plans, code, or post-mortems. Used by architect (pre-impl), pm (planning), senior-dev (impl), l3-support (post-incident).

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

anydesign

無料

Analyze images, websites, and Figma files to extract their design and generate a `design.md` with token system, component inventory, and reconstruction notes. Use this skill whenever the user wants to understand, document, replicate, or audit the design of something visual: a screenshot, a URL, a Figma link, a Pinterest reference, a mockup, a competitor's site, a component, a dashboard, a landing page. Also when they ask 'extract the design system from X', 'document the style of Y', 'analyze this visually', 'convert this image into tokens', 'help me replicate this design', 'what palette does this site use', 'how is this built'. Also for single elements: 'copy this navbar', 'recreate this illustration', 'give me a prompt to regenerate this graphic' — element mode outputs a focused element.md, with token-grounded image-model prompts when the element is visual art. If the user brings any visual source and wants to understand it at a design level — this skill should activate.

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

Shared review framework that every domain reviewer (pci, oracle, gov, edtech, healthcare, mlops, etc.) MUST follow. Defines the output artifact (TM-{slug}.md), mandatory sections, severity scale, verdict format, the workflow scaffold (when-invoked, Step-0 read-inputs, HANDOFF), and the "domain heuristic vs generic check" boundary. Eliminates duplication across the ~30 reviewer prompts.

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

Structured idea generation + multi-LLM debate for the product-owner stage. Diverge (generate genuinely different bets), debate (a 4-persona panel on 4 models argues over 2 rounds), converge (synthesize a recommendation). Used by product-owner before architect; available to architect for design-space exploration.

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

Run the great_cto controlled Codex lifecycle with controller-owned writes, verifier evidence, human gates and optional artifact release.

日本語の概要は準備中です。原文の説明を表示しています。

avelikiy/great_cto1032026年10月12日 更新

avelikiy のスキルをすべて見る

このスキルの問題を報告する