skill creation
needs
Since non-design teams started using Claude, Information Designers started getting more challenging decks to work with, lacking master themes and with extra layers of clutter added. Non-designers need a skill that helps them create presentations that adhere to brand guidelines and are built cleanly, giving designers a starting point to polish. Designers need to see which layouts and components exist, so they can review Claude-made decks faster.
skill 1
Input
Existing PowerPoint decks that don’t match brand guidelines.
Approach
Place the content on top of the existing master template.
Find and update all fonts and colors that don’t match the brand guidelines.
Extract the raw content from slides that are too cluttered to work with.
Output
A restyled PowerPoint deck with the same content.
Status
Released internally. Still used for quick conversions; active development moved to Skill 2.
skill 2
Input
Documents or broad topics for a branded deck.
Approach
Define a deck structure that fits the content and is built using existing layouts.
Confirm the structure with the user before producing any outputs.
Build each slide from the spec file (exact measurements and styles for each layout).
Output
A PowerPoint deck built from scratch with preset layouts.
Status
Beta, tested within the Information Design team. The onboarding documentation is in progress. 18 layouts are available.
The 18 layouts available in beta for Skill 2.
problems & decisions
narrowing skill 1
I tested it mostly on decks from non-design teams. The skill was able to switch the master template but it had issues handling files with more than one master theme referenced — something really common in the raw decks the team gets. The color palette updates worked in simple layouts but with complex ones I started seeing contrast issues and colors applied inconsistently across the page. Since I had to process each deck multiple times to get good results, we decided the skill would be only used for quick conversions and clean-ups.
Decision: We limited the skill’s scope and moved development to Skill 2, where spec files let us control the output.
Conversion Scale
Figma and PowerPoint content is rendered differently because of the way each software measures the total height of a font and defines things like line height and paragraph spacing. The content also looks twice as big when exporting to PowerPoint from Figma directly because of the native resolution of each. Spec calibration is a complex trial-and-error process that takes hours of testing and troubleshooting.
Solution: Build clear conversion rules for content in general and specific rules to handle each family of text styles — paragraphs, captions, headers.
Layout Translation
To reproduce layouts faithfully, Claude needs specific instructions to know where to place each piece of content, as well as clear instructions to build drop shadows, set up the border radius and handle the stacking order correctly. Slides are often built with too many layers and groups, making editing more time-consuming for designers.
Solution: Build Figma components with the fewest layers and groups possible, apply PowerPoint best practices as guidelines in the skill’s instructions, and record each layout’s exact values in its spec file.
Deck generated with the default PowerPoint creation skill.
Deck generated using Skill 2: The branding is applied consistently and the deck doesn't have any generic content that makes it look Claude-made.
— final thoughts
Working on this project has been the best way of getting into AI and learning what it can actually do when designing with and for non-designers. I’ve been polishing my writing skills through prompting — an unexpected win — and I learned lots of things about how each program is natively built, understanding better how to work across them and how to help others navigate them more easily.
I don’t know yet if the team’s vision is something achievable — having a full functioning Figma library that translates to PowerPoint slides with a few prompts — but each technical block so far has been a pivoting opportunity, not an ending.
