Skip to content
Broke2 of 13 blind votes

Does the Stageshow skill improve design in Claude Code? Not in four blind tests

Published · Last checked

Does the Stageshow skill improve design in Claude Code?

No: the Stageshow skill lost to the side without it in all four blind tests in Claude Code, so the verdict is red. Only 2 of the AI judges' 13 votes went to the side with the skill, and that side used more tokens in every attempt, from 1.3x up to 7x. The skill does clean work but doesn't come up with the good idea for you; each time, the side without the skill built the better story and won.

4 experiments, blind votes

ExperimentWhat changedBlind votes (won by the skill side)Token ratio (with skill / without)
Round 1first attempt0/31.3
Round 2general feedback, /stageshow0/32.0
Appeal (Round 3)detailed brief1/36.1
Guide v1small task + brand0/37.1 (incl. subagent)
Guide v4+ 3 critic rounds1/33.8 (total)
Total—2/13—

In Rounds 1 and 2 the voters were me, Hermes (my AI assistant on a server) and Gemini; in the appeal and guide runs, Gemini and two Claude agents with no memory voted; the 2/13 total counts only the AI judges' votes.

Highlights

  • In the first attempt, all three judges found the promo from the side without the skill better; the skill side had no brand and no title, and its water ring never reached the goal, so it lost on the story, not on craft.
  • When both sides got the same feedback sentence and the skill was called explicitly, the result still didn't change: the votes were again 3-0 for the side without the skill.
  • In the appeal, where I wrote out the story point by point, the gap narrowed but the direction didn't change: the votes went 2-1 to the side without the skill, and the skill side used 6.1x the tokens.
  • Even with Benji Taylor's own recipe (a small task, the product's own look, three rounds of feedback), the skill side couldn't win, because it tried to fit a vertical phone into a horizontal card.
  • Asked “which one has the skill,” the AI judges got only 5 of 13 guesses right, which means they assumed the one they liked was the skill version.

How I tested it

I opened two Claude Code sessions in two separate folders: same model (Opus 5.5, effort high), same settings and same prompt, and the only difference was that one had the Stageshow skill installed. The prompt didn't mention the skill by name, so Claude decided on its own whether to use it. The outputs were randomly numbered “1” and “2” and voted on by judges who didn't know which one had the skill. The rule was written before the test: if the skill side didn't get at least two thirds of the votes, the verdict would be red.

The rules every test follows

Who it is for

For anyone thinking about installing a skill in Claude Code, for design or any other work. It shows why you should run a blind comparison on your own work before trusting what a skill promises.

Frequently asked questions

What is the Stageshow skill?

It's a Claude Code skill shared by Benji Taylor, head of design at X. It has Claude make promo visuals such as product screens, short animations or a demo playing on a phone, and it promises less design that looks “made by AI.” I tested version 1.1.1, dated Nov 5, 2026.

What is design slop?

It's what people call it when you ask an AI for an interface or a promo visual and the same thing always comes out: blue or purple gradients, glowing things, badges, cards. It's design that makes you say “an AI made this” at first glance.

Do Claude Code skills actually work?

It depends on the skill. The one I tested, Stageshow, made no difference in four blind tests; the side without the skill won every time. Since new models are already good, some skill rules may no longer make a difference, and Stageshow's author also says some rules may become unnecessary as models improve.

How do I test if a Claude Code skill is worth it?

Open two folders, put the skill in one, and give both the same task. Compare the results without knowing which one has the skill, then look at how long each took and how many tokens it used; in half an hour you'll know whether a skill adds anything for you.