# Comparison prompt

- **[Agent kit](/agent/)**

How it differs from the alternatives: a table that says yes, no or partly in words.

## Prompt

Prompt

> Use the klarhimmel-design skill. Build a comparison of {product} with {alternatives}.
> Install: npx shadcn add @klarhimmel/comparison-table @klarhimmel/comparison-cards (remove what you do not use)
> Pick:
> - comparison-table: a feature-by-feature table, up to 3 alternatives, verdicts in words; the default
> - comparison-cards: who each choice is right for, with an honest closing line
> Props: ours and others as names; rows { feature, note?, values } where values has one yes|no|partial per column, ours first; words { yes, no, partial } are the labels; cards { heading, points }.
> Copy: Be fair: the alternative wins somewhere and the table should show it. Features are things a reader does, in their words. Do not name competitors you cannot compare accurately. Plans against plans belong in the pricing block.
> Accessibility: The table has a caption, real column and row headers, and every verdict is a word as well as an icon. The header stays in view and a caption appears when columns scroll sideways. Check 390 px.
> Never: viewport variants (md:, lg:), style=, text or images outside Text and Media, fields the schema lacks. Shorten copy to fit the limits.
> Done when: every verdict is a word; at least one row where an alternative wins; verdict count matches the columns.

- **Copy for AI**

## Variables

- {product} · What you sell, in a few words · Room booking for small hotels
- {alternatives} · One to three alternatives, fairly · a spreadsheet and a general booking tool

## Blocks it uses

- **[comparison/table](/blocks/comparison/table)**
- **[comparison/cards](/blocks/comparison/cards)**

## A good result

Every verdict is a word. At least one row where an alternative wins. Verdict count matches the columns.

checklist.md

```
- [ ] every verdict is a word
- [ ] at least one row where an alternative wins
- [ ] verdict count matches the columns
```
