The picker is optional, and the interesting part is that skipping it still produces a good deck.
The default has taste
- A dark title slide, then light content slides
- Serif display headings at a confident size
- A three-card row numbered 01, 02, 03, with one accent colour and nothing else
- Generous margins, one idea per slide, and a closing call to action
So the reason to migrate a design system is not rescue. The default will not embarrass you. The reason is brand fidelity: this deck looks like Claude made it, not like your company made it. That is a real argument, and it is a different one from the argument most write-ups will make.
Claude's own summary of that run named the change without being asked: built in the site's navy-and-Satoshi look. Applying a system does not only recolour the deck, it changed the length as well, from five slides to ten.
Docs puts headings on the page within seconds and fills them in while you watch. Slides holds everything back and shows a floating Claude is working label over an empty canvas. An unbranded deck took about two minutes. With a design system applied, the same ask took over six. Both are honest, only one is reassuring.
It is built from files, and driven by a skill
The session panel shows the working: it prints reference files before it draws anything. One run read SKILL.md, craft.md and format.md. Another read fonts.md and deck-files.md. Then it creates files, six for a plain deck and eleven for a branded one, including a deck.json, and publishes them in a single call.
A deck is a set of files, not one opaque object. That is why PowerPoint, HTML and PDF export are all plausible from the same artifact, and it is the same idea as a Claude Code skill: the quality comes from instructions in a file, not from magic in the model.
You know what to expect from a first deck
- Know the default look is defensible
- Know the wait is long and the canvas stays blank
- Understand a deck as files rather than slides
Next: fixing it by pointing at it.