Imported from stantheman0128/stan-claude-harness (skills/design-taste/SKILL.md). Install upstream with npx skills add stantheman0128/stan-claude-harness --skill design-taste. Copyright stays with the author.

Preamble (run first)

_TEL_TIER=$(cat ~/.superstack/config.json 2>/dev/null | grep -o '"telemetryTier": *"[^"]*"' | head -1 | sed 's/.*"telemetryTier": *"//;s/"$//'  || echo "anonymous")
_TEL_TIER="${_TEL_TIER:-anonymous}"
_TEL_PROMPTED=$([ -f ~/.superstack/.telemetry-prompted ] && echo "yes" || echo "no")
_TEL_START=$(date +%s)
_SESSION_ID="$$-$(date +%s)"
mkdir -p ~/.superstack
echo "TELEMETRY: $_TEL_TIER"
echo "TEL_PROMPTED: $_TEL_PROMPTED"
if [ "$_TEL_TIER" != "off" ]; then
_TEL_EVENT='{"skill":"design-taste","phase":"build","event":"started","ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'"}'
echo "$_TEL_EVENT" >> ~/.superstack/telemetry.jsonl 2>/dev/null || true
_CONVEX_URL=$(cat ~/.superstack/config.json 2>/dev/null | grep -o '"convexUrl":"[^"]*"' | head -1 | cut -d'"' -f4 || echo "")
[ -n "$_CONVEX_URL" ] && curl -s -X POST "$_CONVEX_URL/api/mutation" -H "Content-Type: application/json" -d '{"path":"telemetry:track","args":{"skill":"design-taste","phase":"build","status":"success","version":"0.2.0","platform":"'$(uname -s)-$(uname -m)'","timestamp":'$(date +%s)000'}}' >/dev/null 2>&1 &
true
fi

If TEL_PROMPTED is no: Before starting the skill workflow, ask the user about telemetry. Use AskUserQuestion:

Help superstack get better! We track which skills get used and how long they take — no code, no file paths, no PII. Change anytime in ~/.superstack/config.json.

Options:

• A) Sure, help superstack improve (anonymous)
• B) No thanks

If A: run this bash:

echo '{"telemetryTier":"anonymous"}' > ~/.superstack/config.json
_TEL_TIER="anonymous"
touch ~/.superstack/.telemetry-prompted

If B: run this bash:

echo '{"telemetryTier":"off"}' > ~/.superstack/config.json
_TEL_TIER="off"
touch ~/.superstack/.telemetry-prompted

This only happens once. If TEL_PROMPTED is yes, skip this entirely and proceed to the skill workflow.

Wrong skill? See SKILL_ROUTER.md for all available skills.

Design Taste

Design judgment patterns adapted from curated design system references, independent UI craft writing, and observational analysis of high-craft products across SaaS, fintech, trading, and consumer.

The taste layer that sits between brand selection and mechanical implementation. This skill answers: "Given the constraints, what should this actually look like?" It catches generic AI output, makes judgment calls about density and layout, and defines what polished, intentional interfaces feel like.

frontend-design-guidelines tells you how to build it right. brand-design tells you how to brand it. This skill tells you what design direction to take.

Authority Model

• brand-design → palette/typography selection → writes brand.md
• This skill → design direction, judgment calls, anti-slop review
• frontend-design-guidelines → mechanical correctness (a11y, forms, states) → invokes this skill when direction is needed
• page-load-animations → framer-motion implementation recipes

frontend-design-guidelines delegates direction questions TO this skill. This skill does NOT invoke frontend-design-guidelines.

Conflict resolution with brand.md: If brand-design chose Inter as the typeface, this skill does NOT ban Inter. It ensures Inter is used with intentional weight, tracking, and hierarchy — not as a lazy default. Anti-slop rules apply to thoughtless defaults, not deliberate brand choices.

When to Fire This Skill

• The user says "this looks generic" or "this feels AI-generated"
• The user asks what direction a page or component should take
• The user wants something to feel "more premium" or "less generic"
• Building a new page and the aesthetic direction is unclear
• Reviewing existing UI for AI-slop patterns
• The user references a design direction ("make it warm and premium", "dense trading feel")

Do NOT fire on "make this look good", "polish this", or "design this page" — those belong to frontend-design-guidelines.

Workflow

Step 0: Read brand.md if present

Understand the existing aesthetic direction. Respect it — don't override palette or typeface choices.

Step 1: Identify which reference is relevant

• Checking for generic AI patterns → references/anti-ai-slop.md
• Making layout/density/component decisions → references/design-judgment.md
• Looking for style direction and inspiration → references/theme-references.md
• Choosing visual direction for a pitch deck → references/theme-references.md § "Pitch Deck Styles"

Load only what you need. Don't read all references upfront.

Step 2: Apply direction and produce a design brief

When making direction decisions, output a compact design brief:

Direction: [archetype — e.g., "dark dashboard, warm-monochrome style"]
Density:   [spacious / comfortable / compact]
Surface:   [cards on dark bg / flat sections / panels]
Type mood: [3 descriptors — e.g., "tight, technical, mono-heavy"]
Motion:    [1 descriptor — e.g., "crisp springs, no bounce"]
Do:        [3-5 specific direction choices]
Don't:     [3-5 specific anti-patterns to avoid]

Step 3: Run anti-slop check before reporting complete

Scan for the detection signals in anti-ai-slop.md. Flag any that match.

3 Modes

Direction mode

Choosing what the page/component should look like before building. Produces a design brief. Used at the start of UI work.

Review mode

Auditing existing UI for generic/AI-slop patterns. Diagnostic only — detects problems and names them. For constructive polish suggestions, defer to craft-and-polish.md in frontend-design-guidelines.

Quick lookup

"What layout for a dashboard?" → page archetype from design-judgment.md. "What density for a trading surface?" → density guidance. "How should this card feel?" → component direction.

Enforceable Rules (code-scannable)

These can be verified by scanning code. Confidence varies:

1. No pure black #000 or #000000 in product UI — unless it is a deliberate brand token documented in brand.md or theme tokens. Grep for exact hex. (high confidence)
2. No transition: all or transition-all — grep for exact string (high confidence)
3. No arbitrary Tailwind values in spacing — grep for p-\[, m-\[, gap-\[ patterns (medium confidence — some arbitrary values are legitimate)

Judgment Heuristics (review guidance, not hard rules)

These require visual/contextual judgment and cannot be mechanically checked:

1. Single accent + grayscale preferred — multiple competing colors is a smell
2. Content inside cards/panels, not bare page background
3. Entry takes longer than exit (asymmetric timing)
4. Convergence test: "If someone said AI made this, would they believe immediately?"
5. Typography hierarchy: 3 weights max, 4-5 sizes max
6. Whitespace signals confidence — cramped signals cheap
7. Custom easing curves preferred over CSS defaults (hard to grep reliably — review manually)
8. Shadow opacity should be low (4-8% for cards, 12% max for overlays)

State Transition Quick Reference

How states should change (the "feel" layer that states.md doesn't cover):

Resources

references/

• references/anti-ai-slop.md — Detection checklist for generic AI output + fix rules + 11 aesthetic directions
• references/design-judgment.md — Decision trees (cards vs lists, density, asymmetry, form controls, modals), page archetypes, mobile collapse rules
• references/theme-references.md — 6 style references (Warm Monochrome, Stark Minimal, Gradient Trust, Workstation Dense, Soft Consumer, Gallery Editorial) + trait index + mix-and-match composition guide

Cross-skill references

• frontend-design-guidelines/references/craft-and-polish.md — constructive polish (this skill is diagnostic, that skill is constructive)
• frontend-design-guidelines/references/layout-and-design.md — spacing rules and contrast (this skill provides direction, that provides constraints)
• brand-design → brand.md — palette and typography choices (this skill respects them)

Quick Start

# Triggers:
#   "This looks generic / AI-generated"
#   "What direction should this dashboard take?"
#   "Make this feel more premium"
#   "Review for taste"
#   "Make it feel warm-monochrome / stark-minimal"
#   "What density for this trading surface?"

Telemetry (run last)

After the skill workflow completes (success, error, or abort), log the telemetry event. Determine the outcome from the workflow result: success if completed normally, error if it failed, abort if the user interrupted.

Run this bash:

_TEL_END=$(date +%s)
_TEL_DUR=$(( _TEL_END - ${_TEL_START:-$_TEL_END} ))
_TEL_TIER=$(cat ~/.superstack/config.json 2>/dev/null | grep -o '"telemetryTier": *"[^"]*"' | head -1 | sed 's/.*"telemetryTier": *"//;s/"$//' || echo "anonymous")
if [ "$_TEL_TIER" != "off" ]; then
echo '{"skill":"design-taste","phase":"build","event":"completed","outcome":"OUTCOME","duration_s":"'"$_TEL_DUR"'","session":"'"$_SESSION_ID"'","ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","platform":"'$(uname -s)-$(uname -m)'"}' >> ~/.superstack/telemetry.jsonl 2>/dev/null || true
true
fi

Replace OUTCOME with success/error/abort based on the workflow result.