Feedback, Not a Redesign.
The video opens up the design-critique skill, a single plain-language SKILL.md file of about three kilobytes. It explains how the skill is triggered, what it checks, and why it gives feedback instead of redesigning your mockup.
A common assumption is that asking Claude for a design critique means Claude will fix the mockup. The video says it does not. The design-critique skill gives structured feedback on usability, hierarchy and consistency, and it never touches the mockup itself. The video uses the skill to show how a skill works in general.
A folder, not a program
A skill is a folder that Claude reads before it works. This one is design-critique, and it is a single SKILL.md file of about three kilobytes. The file is the whole instruction set, written in plain language with no hidden logic underneath it. Claude reads the file and then acts on what it says. In the video's words, the file is the program.
How the skill is picked up
The skill's job, word for word, is to get structured design feedback on usability, hierarchy and consistency. It triggers when you ask Claude to review a design or critique a mockup, when you ask what it thinks of a screen, or when you share a Figma link or a screenshot. This can happen at any stage, from early exploration to final polish.
Once the skill is triggered, Claude reads the Steps section and runs the steps in order. The process is linear, with no branching unless a step says so.
Feedback, not a redesign
The video stresses two limits. First, the skill hands back feedback and never changes the mockup. Second, it only checks three things: usability, hierarchy and consistency. Anything outside those is outside what it sees.
Because the instructions are fixed, running the same screen through twice returns the same feedback both times. The skill does not improvise a different review each run. It gives structured feedback on those three checks in the same way every time.
Seeing the instructions before the result
To make the instructions visible, the video suggests a particular way of asking. You tell Claude you want structured feedback on the design's usability, hierarchy and consistency, and you ask it to read the design-critique skill and walk you through what it will do before it does it. That final clause matters. Having Claude explain itself first is how you actually see the instructions it is following, not just the result.
Key takeaways
- A design critique from Claude is feedback, not a redesign.
- A skill is a folder of plain-language instructions that Claude reads before it works.
- The design-critique skill is one SKILL.md file of about three kilobytes.
- It triggers on requests to review a design, critique a mockup or look at a screen, including Figma links and screenshots.
- It checks only usability, hierarchy and consistency.
- It runs its steps in order and returns consistent feedback for the same screen.
- Asking Claude to explain what it will do first shows you its instructions.
Try it yourself
Take a mockup you are actually working on and paste this into Claude: you want structured feedback on the design's usability, hierarchy and consistency, and Claude should read the design-critique skill and walk you through what it will do before it does it. Compare the walkthrough with the feedback you get back.
Chapters
Full transcript(auto-generated, with timestamps)
The naive framing: "will Claude fix my mockup?"
[0:00]Someone assumed a design critique means Claude fixes the mock-up itself. It doesn't. It gives structured feedback on usability, hierarchy, and consistency. So, how do I get Claude to critique my design mock-up? A skill is a folder
A folder, not a program
[0:13]Claude reads before it works. This one is design critique, a single skill MD file about 3 kilobytes is the whole instruction set written in plain language with no hidden logic underneath it. Claude reads it then acts on what it says. The file is the program. Its job,
How it gets picked up
[0:28]Word for word, get structured design feedback on usability, hierarchy, and consistency. It triggers when you ask Claude to review a design, critique a mock-up, or say, "What do you think of this screen?" or when you share a Figma link or screenshot at any stage from early exploration to final polish. Once triggered, Claude reads the steps section and runs them in order linear. No branching unless a step says so. That's feedback, not a redesign. The skill reviews usability, hierarchy, and consistency and hands back structured
Feedback, not a redesign
[0:55]Feedback, but it never touches the mock-up itself. Run the same screen through twice and the feedback comes back the same both times, but it only checks those three things: usability, hierarchy, consistency. Anything outside that is outside what it sees. Design critique doesn't redesign your mock-up. It hands back structured feedback on usability, hierarchy, and consistency the same way every time. Your turn. Paste this into Claude. "I want structured feedback on this design's usability, hierarchy, and consistency. Read the design critique
Carry-out
[1:24]Skill and walk me through what you will do before you do it." That clause matters. Having Claude explain itself first before it acts is how you actually see the instructions it's following, not just the result. Feedback, not a redesign. Lay 'em in for bear.





