| name | design-critique |
| description | Run the four checks (swap, squint, signature, token) plus named-bans audit on a design — produce a tabular Before/After/Why critique. Orchestrator that pulls from `design-taste`, `ai-tells-forbidden-patterns`, `motion-and-transitions`, and `pre-delivery-checklist`. Use this skill on any in-progress design artifact (screen, deck, document, social asset, component) before showing it to the user. Output is named-by-rule, not taste-based. |
| type | skill |
| project | skills-library |
| plugin | design |
| aliases | ["design-critique"] |
| tags | ["type/skill","plugin/design","topic/design","topic/critique","scope/orchestrator"] |
| status | active |
Design critique (orchestrator)
Run a craft critique on an existing design. Output is a markdown table — Before / After / Why — that names specific issues by their cataloged ban. Adapted from Emil Kowalski's review-as-table format.
Steps:
-
Confirm the artifact — single screen, single deck, single document, single asset, or component family.
-
Run [[design-taste]]'s four checks:
- Swap test: if I swapped the typeface for the AI default (Inter), would anyone notice? If I swapped the layout for a centered-hero template, would it feel different? Note where swapping wouldn't matter — that's where you defaulted.
- Squint test: blur your eyes. Is hierarchy still perceptible? Is anything jumping out harshly?
- Signature test: can you point to five specific elements where the chosen direction's signature appears? Not "the overall feel" — actual components.
- Token test: read the tokens out loud. Do they sound like they belong to this brief's world, or could they belong to any project?
-
Audit against [[ai-tells-forbidden-patterns]] — cite each violation by its named ban (the LILA ban, the Inter ban, the centered-hero ban, the 3-column-card-row ban, the Jane Doe effect, etc.).
-
For motion-bearing designs, audit against [[motion-and-transitions]] named bans (the scale(0) ban, the ease-in ban, the transition: all ban, the keyframe-on-rapid-trigger ban, the transform-origin-center-on-popovers ban).
-
Run [[pre-delivery-checklist]] CRITICAL and HIGH gates.
-
Output as a Before / After / Why table. No prose-style "Issue 1: ... Issue 2: ..." — tabulate.
Output format:
# Critique: {Artifact} — {Date}
## The four checks
- Swap test: {pass / fail with specifics}
- Squint test: {pass / fail with specifics}
- Signature test: {pass / fail — list 5 elements, or note where signature is invisible}
- Token test: {pass / fail with specifics}
## Findings
| Before | After | Why |
| --- | --- | --- |
| {what's there} | {what to change to} | {named ban or rule + brief rationale} |
| {next finding} | {fix} | {named ban} |
| ...
## CRITICAL items (block ship)
- {Item 1}
- {Item 2}
## Recommended next iteration
- Top 3 fixes by leverage:
1. {fix}
2. {fix}
3. {fix}
- After this iteration, re-run critique before showing the user.
Tone: direct, specific, named. Cite every finding by its rule or ban. Avoid taste-based feedback ("I prefer the second one"). Tie all critique to the brief and the cataloged rules.