A Claude Agent Skill built for an internal UX Center of Excellence: point it at a designer's existing portfolio or bio deck (PDF, PowerPoint, or Keynote export) and it restyles the whole thing into the firm's internal design system, then hands back an annotatable HTML preview so the designer can polish content before exporting a final PPTX or PDF.
Designers at the firm carry portfolio decks built up over years: different templates, different tools, different eras of personal style. Every time the UX CoE needed one presented internally or to a client, someone spent hours by hand rebuilding slides into the current brand system, one shape and font at a time.
Ascend Restyle exists to remove that hours-long manual step. It's not a proposal generator, and it doesn't score candidates against a brief. That's what the separate match / export-proposal skills do. This one is a one-to-one restyle: take a designer's own existing material, keep the content, replace the design.
The skill is deliberately conservative about content. It never invents or rewrites what a slide says, only how it looks. Structure gets confirmed with the designer before any restyling begins.
pptx skill), or a quick in-browser print-to-PDF for a fast pass.This is the firm's design language for internal designer-facing material, distinct from client-facing brand work. The skill enforces its real tokens and a short list of hard rules rather than approximating a "similar-looking" theme.
The internal design system deliberately avoids the soft, rounded-corner look most decks default to. Chamfer-clipped polygons give it a sharper, more technical identity that reads as "internal engineering tool" rather than "marketing template," which matches how the UX CoE wants this material to feel.
The first real test of the skill was my own 49-page Keynote-exported portfolio PDF: Cover, About, Expertise, Experience, Competence Focus, a Portfolio index, seven full case studies, and a Closing section, all restyled in one pass.
Two real content decisions came up mid-restyle: industry tags across my case studies weren't consistently named, so the skill flagged the inconsistency and we standardized the taxonomy together rather than silently restyling around a mislabeled slide.
A later pass rebuilt the layout engine itself around four grid archetypes driven by content density and image count, so dense case-study slides and sparse cover slides don't get forced into the same template. As of this writing, the preview is generated and the annotation pass is still in progress. That's the honest state of a tool built to be used, not just demoed.
Content is the hard part. Restyling forty-nine slides shouldn't take longer than writing them did.