gemini-carousel.

A LinkedIn carousel is one of the few formats where the second-impression rate matters. The reader has to want to swipe, then keep swiping, then complete. gemini-carousel is the skill that builds the spine of a carousel before any image is generated, which is what keeps the multi-slide arc coherent rather than letting each slide drift into a different visual register.

The hard part of a carousel is the spine, which is the through-line of headline ideas across the slide sequence. Most one-shot carousel prompts try to design and write at the same time and produce decks where each slide is locally good and the sequence reads as disconnected. gemini-carousel separates the two passes: the spine first as a per-slide brief, then the image prompts after the writer has approved how the slides relate.

§01What it does

The skill asks for source content (a post, a newsletter section, research notes, or a framework), the brand style (pulled from brand-kit.md if present, typed in, or picked by the model), and the slide count (six, eight, or ten). Then it analyses the content and produces a design brief with three layers: the cover slide hook and visual direction, the body slides each with a headline of eight words or fewer and a body line of fifteen words or fewer, and the closing CTA slide with the repost ask.

The brief is fully editable. The writer reads it and either says generate or asks for specific changes. Once approved, the skill outputs one Gemini image generation prompt per slide, each in its own code block and numbered clearly. Every prompt follows the same structure: brand palette, typography directive, slide context (slide N of M), the headline and body text from the brief, and a per-slide visual element suggestion.

§02The approval gate scales with slide count

The longer the carousel, the more important the gate. A six-slide carousel that goes wrong in one slide is recoverable. A ten-slide carousel where slide four does not connect to slide three breaks the arc and the whole deck has to be rebuilt. Approving the brief slide-by-slide before any image is generated is what keeps the cost of a misfire bounded to one slide rather than the whole deck.

The pattern compounds with gemini-infographic. The brief shape is the same. A writer who wants to test whether their idea reads better as a single infographic or as a multi-slide carousel can run both skills against the same source content and compare the briefs before paying for either generation. The brief artifact is reusable across the two formats.

§03Setup

# Trigger phrases:
#   "carousel"
#   "build a carousel"
#   "turn this into a carousel"
#   "gemini carousel"

Each prompt goes into a Gemini chat with Create Image enabled, generated at 1080×1350. The skill outputs the prompts in numbered order so the writer can paste them sequentially without losing track of the slide sequence.

§04Caveats

Six slides is concise, eight is the standard, ten is a deep-dive. Asking for more than ten dilutes the format; the skill caps at ten, which is the right call.

Slides over fifteen words each lose legibility on the LinkedIn carousel viewer at mobile sizes. The skill enforces the cap; resist the urge to extend during the approval pass for the same reason gemini-infographic enforces its bullet cap.