Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user needs a slide-by-slide carousel brief for Instagram, Facebook, or LinkedIn built on a narrative arc: hook on slide 1, value in the middle, CTA on the last slide. Includes an HTML preview step to approve the concept before production. Trigger on 'carousel brief', 'make a carousel', 'turn this into a carousel', 'slide brief', 'LinkedIn carousel', or 'render a carousel preview'.
.claude/skills/minhnv0807-43-carousel-brief-global/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 131% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 78% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 102% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 162% | 0% |
A carousel is not a stack of images — it is a sequence, and every slide has to pull the reader to the next one. Slide 1 decides whether anyone swipes. The final slide must carry a clear CTA. Never invent the content: if something is missing, ask the user first.
Read .agents/product-marketing-context-global.md, the brand guideline, and the source content if available. If the user has not supplied content, ask up to 4 questions:
A strong carousel has an explicit arc before content is distributed across slides:
Hook -> Problem -> Insight -> Solution -> Proof -> CTAFix the arc first, then split the content. With existing source material: slide 1 is the strongest hook (pain point, question, or surprising number), the middle slides carry proof and explanation, and the last slide carries a clear action.
Each slide gets a headline (max 8 words), body (max 25 words), and visual direction. Copy is approved before any layout work begins.
Never leave the designer guessing. Background colors come from the brand guideline: the hook and CTA slides use a dark or saturated ground (primary color], or a gradient from primary color] to secondary color]) for contrast, while body slides use a light ground for breathing room.
Before handing production to the designer, render one HTML preview file:
The HTML preview approves the concept — it is not a production file. Once approved, the designer builds the real asset in Canva, Figma, or Adobe Express.
File name: carousel-brief-[name]-[YYYYMMDD].md
markdown# Carousel Brief — [Name] ## Overview | Item | Detail | |------|--------| | Purpose | | | Channel | | | Funnel stage | TOFU / MOFU / BOFU | | Slide count | | | Dimensions | 1080x1080 (Feed) / 1080x1350 (Portrait) / 1200x1500 (LinkedIn document) | | Tone | | | Deadline | | | Max revision rounds | | ## Brand direction (applies to the whole carousel) - Dominant ground: [white / primary color / gradient primary to secondary / mixed] - Heading font: [heading font] Bold — color [...] - Body font: [body font] Regular — color [...] - Accent: [secondary color] — used for highlights, figures, and CTA - Logo: at minimum on slide 1 and the final slide (or every slide if the brand guideline requires it), positioned per the guideline ## Slide 1 — HOOK (decides the swipe) | Element | Content | |---------|---------| | Headline | [strong hook — pain, curiosity, or a surprising number; max 8 words] | | Subtext | [one supporting line — max 15 words] | | Visual | [full-color ground / bold typography / emotive image] | | Micro-CTA | "Swipe" — or drop it if the hook is strong enough | | Ground | [dark or saturated — [primary color] — maximum contrast] | ## Slide 2 — PROBLEM / SETUP | Element | Content | |---------|---------| | Headline | | | Body | | | Visual | [icon / illustration / minimal text layout] | | Ground | [light — breathing room after slide 1] | ## Slide 3 — INSIGHT / TURNING POINT | Element | Content | |---------|---------| | Headline | | | Body | | | Visual | [large figure / quote / simple diagram] | | Accent | [highlight the figure in [secondary color]] | ## Slide 4 — SOLUTION / VALUE | Element | Content | |---------|---------| | Headline | | | Body | | | Visual | | ## Slide 5 — PROOF (if available) | Element | Content | |---------|---------| | Headline | | | Body | [testimonial / result figure / short case study] | | Visual | [screenshot / quote card / before-after] | ## Slide N-1 — PRE-CTA / RECAP | Element | Content | |---------|---------| | Headline | [summarize the value in one line] | | Body | [restate the reason to act] | ## Final slide — CTA | Element | Content | |---------|---------| | Headline | [clear CTA — verb plus a specific benefit] | | Subtext | [what happens next] | | CTA button | [text plus link, or "comment [keyword]"] | | Logo + contact | [full logo, website, handle, or email per campaign] | | Ground | [dark ground matching slide 1 — closes the visual loop] | ## Do not - Use colors outside the brand palette - Overload the frame with text — keep overlay text light and legible - Substitute fonts other than [heading font] / [body font] - Put more than one idea on a slide
| Channel | Format | Notes | |---------|--------|-------| | Instagram / Facebook | 1080x1080 (1:1) or 1080x1350 (4:5) | 2-10 cards; 4:5 occupies more of the mobile screen | | LinkedIn document post | 1080x1080 or 1200x1500, exported as PDF | Uploaded as a document; keep type larger — it renders small in feed | | Pinterest idea pin | 1000x1500 (2:3) | Vertical only; heavier text tolerance than Meta |
01-content-calendar-global: where the carousel sits in the publishing calendar.42-image-brief-global: use for a single static instead of a slide sequence.41-campaign-asset-list-global: the carousel's place in the campaign asset list and its deadline.45-design-tool-brief-global: production direction once the concept is approved.46-brand-guideline-global: source of colors, fonts, and logo rules.47-design-review-global: score the carousel after submission.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | fail→fail | 18,351 | 17,883 | -3% | 1 | 1 | 0% | 2,876 | 5,348 | +86% | 0 | 0 | — |
case-07 | fail→fail | 15,702 | 14,588 | -7% | 1 | 1 | 0% | 2,576 | 3,881 | +51% | 0 | 0 | — |
case-01 | fail→fail | 29,791 | 37,579 | +26% | 1 | 1 | 0% | 4,382 | 7,449 | +70% | 0 | 0 | — |
case-02 | fail→fail | 32,720 | 19,307 | -41% | 1 | 1 | 0% | 4,477 | 6,052 | +35% | 0 | 0 | — |
case-03 | fail→fail | 32,760 | 25,888 | -21% | 1 | 1 | 0% | 5,400 | 7,702 | +43% | 0 | 0 | — |
case-04 | fail→fail | 24,660 | 8,931 | -64% | 1 | 1 | 0% | 3,282 | 3,777 | +15% | 0 | 0 | — |
case-05 | fail→pass | 8,938 | 12,269 | +37% | 1 | 1 | 0% | 1,470 | 3,403 | +131% | 0 | 0 | — |
case-06 | fail→pass | 11,356 | 12,249 | +8% | 1 | 1 | 0% | 1,951 | 3,467 | +78% | 0 | 0 | — |
case-09 | fail→pass | 12,085 | 10,162 | -16% | 1 | 1 | 0% | 2,009 | 4,064 | +102% | 0 | 0 | — |
case-10 | pass→pass | 11,302 | 10,258 | -9% | 1 | 1 | 0% | 1,848 | 4,007 | +117% | 0 | 0 | — |
case-11 | pass→pass | 11,773 | 10,702 | -9% | 1 | 1 | 0% | 2,171 | 3,977 | +83% | 0 | 0 | — |
case-12 | fail→pass | 12,883 | 11,003 | -15% | 1 | 1 | 0% | 2,106 | 4,150 | +97% | 0 | 0 | — |
case-13 | fail→pass | 7,178 | 5,268 | -27% | 1 | 1 | 0% | 1,209 | 3,173 | +162% | 0 | 0 | — |
case-14 | fail→pass | 10,768 | 8,812 | -18% | 1 | 1 | 0% | 1,860 | 3,847 | +107% | 0 | 0 | — |
case-15 | fail→pass | 8,520 | 3,315 | -61% | 1 | 1 | 0% | 1,462 | 2,801 | +92% | 0 | 0 | — |
case-16 | fail→fail | 11,104 | 19,112 | +72% | 1 | 1 | 0% | 1,981 | 6,024 | +204% | 0 | 0 | — |
case-17 | pass→pass | 9,250 | 8,056 | -13% | 1 | 1 | 0% | 1,641 | 3,579 | +118% | 0 | 0 | — |
case-18 | fail→pass | 11,024 | 11,058 | +0% | 1 | 1 | 0% | 1,830 | 4,086 | +123% | 0 | 0 | — |
case-19 | fail→pass | 31,142 | 4,601 | -85% | 1 | 1 | 0% | 1,337 | 3,086 | +131% | 0 | 0 | — |
case-20 | pass→pass | 13,689 | 7,251 | -47% | 1 | 1 | 0% | 2,294 | 3,461 | +51% | 0 | 0 | — |
case-21 | fail→fail | 3,588 | 6,482 | +81% | 1 | 1 | 0% | 542 | 3,346 | +517% | 0 | 0 | — |
case-22 | pass→pass | 14,957 | 19,785 | +32% | 1 | 1 | 0% | 2,920 | 5,763 | +97% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +41 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.