This one's for when you need to actually see what your PDFs look like, not just extract text and hope for the best. It leans hard on visual validation by rendering pages to PNG with Poppler's pdftoppm, then uses reportlab for generation and pdfplumber for extraction. The workflow is deliberate: create, render, inspect, fix, repeat. It's opinionated about quality too, explicitly calling out common disasters like clipped text, overlapping elements, and those mysterious black squares that show up when fonts go wrong. If you've ever shipped a PDF that looked fine in code but broken in practice, this skill is built around preventing exactly that. Good fit for reports, invoices, or anything where layout isn't negotiable.
npx skills add https://github.com/openai/skills --skill pdf