agentsclimarketplace

Aesthetic coherence

Skill pixelcrafts-app/agent-skills/skills/web/aesthetic-coherence

Detect aesthetic mixing — a screen committing to two design languages at once (glass + neumorphism, bento + brutalist) is the #1 "assembled, not designed" tell. Scores each aesthetic's signals, flags mixed signatures, then hands per-aesthetic spec compliance to the audit engine. Use on a surface that feels "off" but passes token/contrast audits.From its SKILL.md

Install
npx -y skills add pixelcrafts-app/agent-skills --skill aesthetic-coherence

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.6 KB, as published. Nobody here has run it

Aesthetic Coherence Audit (web)

Detection is a signal-scoring pass (classification, not compliance). Spec compliance is delegated to the engine (design:aesthetic-coherence).

Step 0 — Inputs

Read $ARGUMENTS (file/page/app; if app, enumerate routes under app/ or pages/). Read design-tokens.md for a declared aesthetic.

Step 1 — Score each aesthetic's fingerprint per file

  • Minimalist — no backdrop-filter, ≤2 shadow levels, ≤2 brand colors, high whitespace, small radius, no decorative gradients.
  • Flat — zero shadows, solid fills, 1px token borders, no effects.
  • Material — ≥3 shadow levels, elevation, ripple motion, FAB.
  • Utility-brutalist (Linear/Vercel) — font-mono metadata, rule-heavy tables, dark-first, single accent, minimal motion, tinted neutrals.
  • Glassmorphism — backdrop-filter: blur, bg alpha .1–.3, hairline alpha borders, layered translucency.
  • Neumorphism — dual offset shadows on same-surface color, no border/fill, interactive elements.
  • Claymorphism — radius >16px, soft-saturated fills, inner+outer shadow, playful vocabulary.
  • Liquid Glass (iOS 26) — backdrop-filter + animated specular, spring motion, morphing translucent chrome.
  • Bento — grid with varying col/row-span, uniform radius 16–24, per-tile hover, ≤4 tile sizes.
  • Editorial — type ratio ≥1.5, serif/two-sans pair, measure 55–70ch, visible grid.
  • Brutalist — pure #000/#fff, system/mono fonts, radius 0, no shadow, asymmetry.
  • Dark-cinematic — true/near-black bg, graded color, glow/bloom, image/video-forward.
  • AI-native — animated mesh gradients, 3D orbs, particle motion, dark + bright bloom.
  • Retro/Y2K — chrome gradients, pixel fonts, skeuomorphic icons.

Step 2 — Classify per file

Take the top-2 scores. Large gap → COMMITTED (healthy). Gap ≤30% → MIXED (FAIL vs single-aesthetic rule). No strong signal → UNCLEAR (not necessarily a fail). Output dominant/secondary/gap/verdict + file:line evidence per signal.

Step 3 — Cross-file coherence

Aggregate per-file verdicts; flag outliers committed to a different aesthetic than the app. Cross-file mixing (a glass modal in a brutalist app) is worse than in-file and is a FAIL.

Step 4 — Delegate compliance (once dominant aesthetic is known)

verify-changes brief:
  scope: <COMMITTED/MIXED files>
  dimensions: [design:aesthetic-coherence]
  depth: direct
  fix: no
  context: { aesthetic: <dominant or user-confirmed> }

Aesthetic choice is user taste — if UNCLEAR or conflicting, ask before delegating. Never run compliance against a guessed aesthetic.

Step 5 — Report

Detection (app aesthetic vs declared, mixed files, outliers) + engine's per-rule PASS/FAIL. Verdict: 0 mixed/outliers/fails → COHERENT · any mixed/outlier → FRAGMENTED · only rule fails → COMMITTED-BUT-UNFINISHED.

Step 6 — Fix loop (never automatic)

Aesthetic fixes are high-blast-radius. Propose which aesthetic to keep (strongest signal default), list exact per-file changes, wait for confirmation per file, apply only confirmed, re-run Steps 1–3. Never fix: yes — taste is not a rule-driven auto-fix.

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.