Using AI Agents to Build and Review Slide Decks

You cannot hand a model a .pptx and ask it to fix slide 12. Not usefully. The model would have to either drive a GUI, which is slow and brittle, or edit raw OOXML, which is a format nobody should be editing by hand let alone asking a language model to patch blind. Either way you’re not getting a reliable result.

Hand it a text file instead and the job is just text editing, which is the thing these models are actually good at.

What’s actually wired up

Slaide’s core ships an MCP server and an Agent Skill, both Apache-2.0, both meant for exactly this. Claude Code, Cursor, Codex, anything that speaks MCP or reads a skill file can generate and edit .slaide files directly.

A prompt I actually use, roughly:

Write a five-slide deck summarizing this quarter’s numbers: MRR growth, churn, new customers, the two product changes that shipped, and next quarter’s one priority. Keep it to one idea per slide. Use the ink theme.

What comes back is a .slaide file, Markdown headings for slides, short lines under each, the YAML theme block already filled in. I render it, look at the PDF, and usually ask for one fix: a slide that’s too dense, a number that needs a source, a headline that’s stating the obvious instead of the actual point.

slaide render quarterly.slaide --out quarterly.pdf

That loop, prompt, render, look, adjust, takes a few minutes for a deck that would’ve taken half an hour to type into PowerPoint by hand.

Review works the same way

Because the deck is text, “review this deck” is a normal code review prompt, not a special case. I can ask an agent to check a .slaide file against a style guide, flag any slide with more than three lines, or confirm every number in the deck has a citation in the notes. It reads the file, it doesn’t need a screenshot, it doesn’t need to render anything first. That’s a real advantage over asking a model to review a PDF or a set of slide images, where it’s inferring content from pixels instead of just reading the words that are there.

The honest limit

The agent writes text. It does not design. If you ask it for a “clean, professional deck” it will write clean, professional sentences and apply whichever theme you told it to use. It will not invent a new layout, choose typography, or make a judgment call about whitespace, because none of that lives in the text file it’s editing. The theme system is deliberately simple for this reason, a palette and a font, not a full design language, so there’s a small, well-defined surface for the model to work within instead of an open-ended design problem it isn’t equipped to solve.

That’s a feature for the decks this is aimed at, status reports, customer summaries, internal updates, anything where the content is the deliverable. It’s a real limit for anything where the design is the deliverable. Don’t ask an agent to build you a keynote that needs to look like it came out of a studio. Ask it to write the ten sentences that are actually in your head, correctly, in the right order, and let the renderer handle the part that doesn’t need a design opinion.

I use this mostly for decks where I know what I want to say and the friction was always typing it into a GUI. Removing that friction is the whole value. It’s not making decisions I wouldn’t have made, it’s just not making me click through a ribbon menu to write them down.