LinkedIn Creative

What it does

The skill builds each creative as an HTML page and renders it to PNG or PDF with headless Google Chrome. No Figma, no MCP servers. The brand is a separate JSON file: colours, fonts, title style, rubric chip, footer. Mine is the default. For your company you copy the empty template and fill it from your brand book or site CSS; the guide ships with the skill. The rules are the same for every brand. The title names the thing; the post's first line stays in the post. Where the data has a shape, it gets a chart instead of text in boxes. Six to nine blocks, the takeaway sits inside the image, and every number has a source. After rendering, the skill checks the layout itself: did the webfonts load (so Cyrillic does not fall back to a serif), does anything stick out or hide under the footer, are there empty bands. Free, MIT licence.

Input and output

Key features

• 24 visual shapes, from a comparison table to a tier list; each says which content it fits and how it usually gets ruined
• Your own brand in one JSON file: colours, fonts, title, footer
• Ready charts: donut with up to 4 slices, bars, a funnel with conversion between stages
• Carousel on a fixed structure: cover, one thought per slide with a visual, a single big-number slide, a closing call to action
• Layout check after rendering plus a list of items to tick before you publish
• Needs Python 3.9+ and Google Chrome on your machine, easiest in Claude Code; no Figma

How to install

Check it works: type /linkedin-creative in the chat — the skill should answer. If it does not: skills are read at session start, so restart Claude.