agentleFS
alexuraxinalexuraxin/retainer/skills/frontend-taste/SKILL.md

frontend-taste

Skill3 starsChanged 3 days agoMITView on GitHub

Call BEFORE the first line of markup when creating or redesigning any HTML interface, page, report or layout. Trigger words (RU+EN): макет, набросаем, вёрстка, страница, экран, дизайн, редизайн, CSS, шаблон, лендинг, отчёт-страница / mockup, layout, page, screen, design, redesign, CSS, template, landing, report page. Audit against generic-AI patterns: typography, font weight, color and surfaces, layout, states. Do not skip because «the change is small» or «I'll take the style from the project» — the observed failure is exactly that. Complements dataviz with a layout pass.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Default fonts / Inter everywhere → a font with character (Geist, Outfit, Satoshi.

    line 17
  2. Distinctive:

    Only 400/700 → add 500/600 for fine hierarchy.

    line 21
  3. Distinctive:

    No max-width → a container ~1200–1440px with auto margins.

    line 40
  4. Distinctive:

    No hover → a background shift/slight scale/translate.

    line 48
  5. Very specific:

    No active/pressed → scale(0.98)/translateY(1px).

    line 49
  6. Distinctive:

    No focus ring → a visible focus (accessibility, not optional).

    line 51
  7. Very specific:

    Only HTML/UI markup and visual artifacts.

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Taste in markup

Turn dull generated output into a considered design. Works with any CSS (Tailwind, vanilla, styled-components). Do not rewrite from scratch — improve what exists. Order: Scan (stack, styling method) → Diagnose (by the audit below) → Fix (surgically, within the existing stack).

Complements dataviz (charts/palette/KPI) with a layout and typography layer — they are about different things and do not conflict.

Typography

  • Default fonts / Inter everywhere → a font with character (Geist, Outfit, Satoshi; for editorial — a serif heading + sans body).
  • Headings with no weight → larger display, tighter tracking (negative), smaller line-height.
  • Too-wide text → ~65 characters per line, more line-height.
  • Only 400/700 → add 500/600 for fine hierarchy.
  • Numbers in a proportional font → font-variant-numeric: tabular-nums for data.
  • Orphans (one word on the last line) → text-wrap: balance/pretty.

Color and surfaces

  • Pure #000000 background → off-black (#0a0a0a, #121212, a dark navy).
  • Oversaturated accents → saturation < 80%, desaturate. One accent, not several.
  • A mix of warm and cool grays → one gray family, a single hue.
  • The "AI gradient" purple-blue → a neutral base + one considered accent.
  • A flat box-shadow → tint it to the background, a single light direction.
  • Empty flat sections → a background image/pattern/low-opacity ambient gradient.
  • A sudden dark section on a light page (or vice versa) → hold a single tone, do not jump to #111.

Layout

  • Everything centered/symmetric → break it: offset margins, left alignment, asymmetry.
  • Three equal feature columns → a 2-column zigzag, an asymmetric grid, masonry.
  • height: 100vh → min-height: 100dvh (iOS Safari).
  • No max-width → a container ~1200–1440px with auto margins.
  • The same border-radius on everything → vary it (inner ones tighter).
  • Buttons not aligned to the bottom of cards → pin to the bottom (a shared horizontal).
  • Feature lists start at different heights → align the baselines.
  • Too little air → double the spacing (for marketing; data dashboards are denser).

States and interaction

  • No hover → a background shift/slight scale/translate.
  • No active/pressed → scale(0.98)/translateY(1px).
  • Instant transitions → 200–300ms on interactive elements.
  • No focus ring → a visible focus (accessibility, not optional).
  • Spinners → skeleton loaders in the shape of the layout.
  • An empty state → a composed "getting started", not emptiness.

Banlist: clichés, placeholders, generic defaults

The audit above is about form; here are concrete bans on content and defaults by which markup reads as "AI-generated".

AI clichés in UI text: Elevate, Seamless, Unleash, Next-Gen, Game-changer, Delve, "supercharge your workflow", "take it to the next level" → flat concrete language for what the feature REALLY does. The same "no neuroslop" rule applied to interface copy, not to the answer's prose.

Filler placeholders: John Doe / Jane Smith / Acme Corp / Lorem Ipsum / user@example.com / "Your Company Here" → realistic contextual content for the domain (names, companies, amounts, dates that could be real). Filler betrays unfinished work.

Generic defaults (by default — NOT these):

  • Fonts: Inter / Roboto / Open Sans / the system sans → a font with character (see Typography).
  • Icons: thin single-line libraries (Lucide, Feather, Heroicons) by default → match the weight and tone of the layout, not the first library at hand.
  • Shadows: heavy default shadow-md/lg/xl → lighten and tint them (see Color and surfaces).
  • Emoji in markup / alt / as placeholder icons → a real icon or nothing.
  • Effects: gradient / neon / 3D glassmorphism by default → extends the "AI gradient" ban to neon and glassmorphism (see Color and surfaces).

When NOT to apply

Only HTML/UI markup and visual artifacts. Backend/CLI/data/logic — not here (that is gate/code-reviewer/graph).

Related: [[task-cycle]].

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.