sxsphinxstack

Skills / Design critique

Design critique

Critique something the person designed, then fix it with them. Names what works and what fails, and applies the top three fixes. Use when they say "is this good", "review my site", "why does this look off", or want feedback before sharing.

Use this skill. Nothing to install.

Open in ChatGPT Open in Claude prefilled and ready; just hit send
for Gemini, Copilot, or Cursor: paste it, then say "use the design critique skill"

Don't have an agent?  ·  Raw file: skills/design-critique.md

Watch the first two minutes

This is how a session goes, including the part where it stings a little. Click through it.

​(pastes the skill) roast my portfolio site. i can't tell if it's good or just done
Critique, yes; roast, no. Every criticism you get will name the element, the reason, and the fix, because "it looks off" teaches nothing. First: what is the site for, and who is it for? I judge against your goal.
getting web dev gigs from small businesses. here's the url
Then the reader is a busy owner, probably on a phone, so I'm looking at it at phone size. What works: your project titles say what each thing does, the load is fast, and there's exactly one page, which owners prefer. Want the failures ranked?
give it to me straight
Six items, ordered by impact for effort. Top: reading order. The eye lands on your photo, but "I build sites for local businesses" sits at the bottom in the smallest type. Second: four fonts are fighting. Third: the gray-on-gray project titles disappear on a phone in daylight. Four through six are smaller.
ouch, ok. the photo stays though
It can stay; the fix changes what the eye does after the photo. We apply the top three now, each shown before and after, your call on every one. Items four to six get saved as a note in the repo for next time. Then you re-critique the updated site in your own words, and we see if the patterns transferred.

Scripted example of a real session.

I just want to know if it's good

A verdict wouldn't help you, because good is relative to a goal: a loud poster for a school fair is right to be loud. So the skill asks what the thing is for and who will see it, then judges against that, at the size the real audience sees it. What you get is capped at six failures, because a twenty-item critique flattens into noise, and the list comes ranked by impact for effort. The ranking is the judgment, and it is the part of a designer's eye you are borrowing until you grow your own.

What you end up with

A written critique, the top three fixes already live, and the rest saved. Here is the writeup for the portfolio above.

Example
CRITIQUE: PORTFOLIO SITE
Goal: web dev gigs from small businesses · Reviewed at phone size
WHAT WORKS 1
Project titles state what each project does. Page loads fast on mobile. One page total, which suits a busy reader.
WHAT FAILS AND WHY 2
1. Reading order. The photo wins the first glance; the offer line is bottom-right at the smallest size. Cost: an owner leaves without learning what you sell.
2. Consistency. Four fonts across one page. Cost: reads as unfinished, which is fatal for someone selling web work.
3. Contrast. Project titles are #999 on #EEE. Cost: invisible on a phone outdoors.
4. Two different corner radii on cards. 5. Uneven gaps between sections. 6. Decorative divider doing no work.
APPLIED TODAY 3
Fixes 1 to 3, each shown before and after, owner's call on each. The photo stayed; the offer line moved to the top at headline size. Site redeployed.
SAVED FOR NEXT SESSION 4
Items 4 to 6, as notes/critique.md in the repo. Owner's own re-critique afterward: "the gaps are what I'd fix next." That matches item 5.
  1. Strengths are specific. "Titles state what each project does" is repeatable on the next project. Vague praise gets you nothing; invented praise is banned.
  2. Every failure has three parts. Element, principle, cost. That structure is what makes a criticism teachable instead of just accurate.
  3. Talk becomes edits. The top three fixes went live in the same sitting, which is the difference between a critique and an opinion.
  4. The loop closes. The owner re-critiqued in their own words and independently found item 5. That is the eye getting sharper, on record.

Questions people actually ask

Is this free?

The skill is free, no account, no catch. The agent you paste it into may or may not be: Gemini has the widest free tier, Copilot's free tier is enough to try it, ChatGPT and Claude free plans can also run it in a chat.

Will it just be nice to me?

No. The skill's rule is that praise must be real: two or three specific strengths when they exist, and when little works, one true thing before moving on. It never invents compliments, because you would learn to distrust everything else it says.

My design is intentionally loud. Will it get marked down?

No, because critique happens against your goal. The skill asks what the piece is for and who it is for before judging anything, and it says explicitly that a loud poster for a school fair is right to be loud. Loudness that serves the goal is a strength; loudness that buries the date is a failure.

What if I disagree with a fix?

You try it both ways. The skill applies its version, you keep yours alongside, and you compare the two. Your taste decides, and whichever way it goes, the reason gets stated in one line so the pattern is learnable.

What can I bring to be critiqued?

Anything you actually designed: a site, poster, deck, profile, or logo, as a URL, file, or screenshot. It gets reviewed at the size the audience sees it, including on a phone if it is a web thing.

Where to go from here

If the same failures keep appearing:

Design basics Brand kit Pick fonts and colors

Make the next thing worth critiquing:

Portfolio page Slide deck Poster or flyer
Curious? Read the full skill — the exact instructions your agent gets 
---
name: design-critique
category: design
description: Critique something the person designed, then fix it with them. Names what works and what fails, and applies the top three fixes. Use when they say "is this good", "review my site", "why does this look off", or want feedback before sharing.
---

# design-critique

Critique something the person actually designed — a site,
poster, deck, profile, logo. The goal is a sharper eye, and the
proof is their artifact getting better in the same sitting. Honest
but specific: every criticism names the element, the reason, and
the fix. "It looks amateurish" teaches nothing; "three fonts are
competing, reduce them to one" teaches the pattern.

## Before judging

Ask what it is for and who it is for — critique against their goal,
not an imagined one. A loud poster for a school fair is right to be
loud. Then look at the real thing (URL, file, or screenshot) at the
size the audience sees it, including on a phone if it is a web
thing.

## The critique

Deliver it in three parts, in this order:

1. What works — 2–3 real strengths, specific enough to repeat on
   the next project ("the headline size makes reading order
   obvious"). Never invent praise; if little works, say one true
   thing and move on.
2. What fails and why — each item as element, principle, cost. Walk
   the usual suspects: reading order (what does the eye hit first,
   is that right?), spacing (crowded or uneven gaps?), alignment
   (how many edges exist?), contrast (does any text disappear?),
   consistency (how many fonts, colors, corner radii?), honesty
   (any decoration doing no work?).
3. Prioritized fixes — the full list ordered by impact for effort,
   numbered. Usually the top item is hierarchy or spacing; a font
   swap almost never leads.

Keep it to at most six failures. A twenty-item critique flattens
into noise; the ranked list is where the judgment lives, and
ranking is the skill they are borrowing.

## Apply the top three

Take the top three fixes and apply
them to the real artifact one at a time, showing before and after
for each. If they disagree with a fix, try their version too and
compare — their taste decides, and either way the reason gets
stated in one line. Ship the improved artifact in its real place:
redeploy, re-export, or commit.

## Close the loop

- Have them re-critique the updated artifact in their own words:
  what would they fix next? Their answer shows whether the patterns
  transferred.
- Leave the remaining fix list (items four onward) as a short note
  in the project so it is there next session.

## Done

- Written critique delivered with specific evidence and a ranked fix list
- Top three accepted fixes live in the artifact, with before/after evidence
- Remaining work saved and the person can explain the next fix in their words

If the same failures keep
appearing across their work, suggest design-basics to drill the
fundamentals, or brand-kit if the failures are consistency ones.