ai-smell-detector

Use to audit a generated or heavily templated interface for generic AI aesthetics, unjustified trend patterns, component-registry residue, vague copy, repetitive motion, novelty theater, or visual dec

SKILL.md

GitHub

AI Smell Detector

Detect patterns, not forbidden shapes

The following are smells when repeated without a project-specific reason:

  • centered hero + generic eyebrow + two CTAs;
  • arbitrary bento/three-card feature grids;
  • gradient text and blue/purple glow as default “premium” styling;
  • glassmorphism with no material concept;
  • excessive pills/badges;
  • identical corner radius on every surface;
  • icon + heading + paragraph card repetition;
  • every section using the same fade-up reveal;
  • oversized vague claims with no evidence;
  • stock imagery with inconsistent lighting/crop;
  • generic font pairing with no typographic character;
  • floating decorative objects with no composition role;
  • 3D cursor/particles/parallax that communicate nothing.

Test each suspicious decision

Ask:

  1. What user/brand/content purpose does this serve?
  2. Would the design lose meaning if it were removed?
  3. Is this repeated because of a system or because it was an easy default?
  4. Does the choice appear in the references as a principle or as a copied fashion cue?
  5. Is there a more specific expression of the same intent?

Output

Classify each smell as: keep, adapt, replace, or remove, with the reason. Avoid redesigning merely to be unusual.