Design Taste
Overview
The judgment layer between building and code review. Given a UI, mock, screenshot, or description, decide what works, what fails, and what to fix first. It never builds components and never grades code.
When to use
- Critiquing a UI, mock, or screenshot description.
- Type, spacing, and palette decisions with contrast math.
- A Nielsen heuristic pass with evidence and severity.
- Diagnosing default-looking AI output and fixing it.
- Comparable 0-10 scores across rounds; a taste practice loop.
When not to use
- Building components, tokens, or WCAG conformance work: pick
ac-frontend-ui.
- Reviewing the code behind a UI: pick
ac-code-review.
Subskill index
| subskill file |
when to load |
subskills/critique-protocol.md |
Perception order, the 10-minute checklist, hosting a crit |
subskills/typography.md |
Type scale, pairing, measure, hierarchy without color |
subskills/spacing.md |
The 4/8 point grid, optical alignment, density budgets |
subskills/color-contrast.md |
Palette shape, 4.5:1 and 3:1 ratios, colorblind safety |
subskills/heuristic-evaluation.md |
Nielsen's ten run as a protocol with severity and evidence |
subskills/ai-slop-anti-patterns.md |
Default-LLM visual tells, diagnosed and fixed |
subskills/rubric-scoring.md |
Anchored 0-10 scores, caps, and round deltas |
subskills/taste-development.md |
Practice reps: teardowns, constraint drills, canon, peer review |
1---2name: ac-design-taste3description: "Judging and critiquing UI and visual-design quality: the judgment layer between building an interface and reviewing its code. Eight leaves: the critique protocol and perception order; typography scale and pairing; spacing grids and optical alignment; color and WCAG contrast math; Nielsen heuristic evaluation run as a protocol; AI-slop anti-patterns; a 0-10 scoring rubric; a taste-development practice loop. Use when asking \"critique this UI\", \"does this design look right\", \"score this mock\", \"is this contrast acceptable\", \"run a heuristic pass\", \"this UI looks like default AI slop\", \"sharpen my design taste\". Not for: building components (use ac-frontend-ui) or WCAG conformance checklists (use ac-frontend-ui)."4license: MIT5metadata:6 source: https://skillmd.com/skills/jiayaoqijia/ac-design-taste7 domain: software-engineering8---910# Design Taste1112## Overview1314The judgment layer between building and code review. Given a UI, mock, screenshot, or description, decide what works, what fails, and what to fix first. It never builds components and never grades code.1516## When to use1718- Critiquing a UI, mock, or screenshot description.19- Type, spacing, and palette decisions with contrast math.20- A Nielsen heuristic pass with evidence and severity.21- Diagnosing default-looking AI output and fixing it.22- Comparable 0-10 scores across rounds; a taste practice loop.2324## When not to use2526- Building components, tokens, or WCAG conformance work: pick `ac-frontend-ui`.27- Reviewing the code behind a UI: pick `ac-code-review`.2829## Subskill index3031| subskill file | when to load |32|---|---|33| `subskills/critique-protocol.md` | Perception order, the 10-minute checklist, hosting a crit |34| `subskills/typography.md` | Type scale, pairing, measure, hierarchy without color |35| `subskills/spacing.md` | The 4/8 point grid, optical alignment, density budgets |36| `subskills/color-contrast.md` | Palette shape, 4.5:1 and 3:1 ratios, colorblind safety |37| `subskills/heuristic-evaluation.md` | Nielsen's ten run as a protocol with severity and evidence |38| `subskills/ai-slop-anti-patterns.md` | Default-LLM visual tells, diagnosed and fixed |39| `subskills/rubric-scoring.md` | Anchored 0-10 scores, caps, and round deltas |40| `subskills/taste-development.md` | Practice reps: teardowns, constraint drills, canon, peer review |