| name | artifact-linkedin-carousel |
| description | Use when asked to build a LinkedIn carousel or document post. Generates slide-by-slide copy, a design-generation prompt for the PDF, and the accompanying post copy from the messaging canon and visual identity. |
Artifact: LinkedIn Carousel
Renders a carousel draft from approved context: a designed multi-slide document at 1080x1350,
in the brand's colors and type, plus the accompanying post copy. A human reviews it before it
ships. See rendering.md for the build and QA mechanics.
What this skill produces
A designed carousel rendered to a 1080x1350 .pptx (exported to PDF for the LinkedIn
document post when LibreOffice is available), in the visual identity, plus the post copy in a
companion .md. A draft for human review.
Reads from context
From the canon, by component ID:
themes, topics: the carousel must sit under an existing theme; if the requested topic matches none, say so before writing
point-of-view: the thesis a thought-leadership carousel argues
customer-proof, key-metrics: evidence slides, where the argument needs them
icp (job to be done, objections): the customer's framing of the problem, in their words
And the full visual identity for the render, plus
brand-writing-identity: coined terms to use
correctly, banned terms to avoid, and the voice rules.
Asks the human
One at a time, and only what context cannot hold:
- Topic and thesis. What is the carousel arguing? One sentence.
- Slide count. 10 is the sweet spot; 8 to 12 acceptable.
- Ending. CTA, recap, checklist, or question? (See the ending patterns in carousel-patterns.md.) If there is no concrete offer, do not fake a CTA; recommend recap or checklist.
Workflow
- Read the canon components and visual identity listed above. Confirm the topic sits under a canon theme.
- Ask the three questions.
- Pick a hook formula from carousel-patterns.md that fits the thesis, and draft slide 1 to carry the whole argument.
- Plan the remaining slides: one idea per slide, within the density limits in the reference file, at least 3 led by a visual element rather than text.
- Draft the post copy: hook line matching slide 1, expansion, preview, and the interaction CTA. This goes in the companion
.md.
- Render the carousel per rendering.md: a
pptxgenjs build on a 1080x1350 custom layout, visual identity mapped to color constants and brand fonts, one slide per plan row, the brand motif repeated. Export to PDF when LibreOffice is available.
- Run content and visual QA per rendering.md, then the checklist, and hand back the file path.
Checklist
- Slide 1 carries the thesis; no warm-up slides
- One idea per slide; headline 10 words or fewer, body 35 words or fewer
- No banned terms from the brand writing identity; no em/en dashes survive the content QA
- At least 3 slides are visual-led
- Slide structures vary; not every slide is headline-plus-body
- The post's first line matches slide 1 (the hook match rule)
- Post is 100 to 250 words, hook inside the first 150 characters, 3 to 5 topical hashtags at the bottom
- All claims trace to canon language or cited evidence; nothing invented
- The rendered carousel is 1080x1350, on-brand, and QA'd (visual pass run, or its absence stated in the handoff)