pdflisted

# Ryu PDF Use this skill for PDF work. Keep the source document and generated output in the user's requested workspace, preserve the original input, and report the exact output path. Ryu bundles this skill, so its instructions are available offline. ## Choose the right path - **Designed React/TypeScript document**: use [pdfcn](https://www.pdfcn.dev/) components with Takumi or Forme. This is the preferred path for branded invoices, reports, statements, and security reports. - **Python-only or coordinate-heavy document**: use ReportLab. - **Existing PDF edits**: use pypdf or pdf-lib for merging, splitting, rotation, metadata, encryption, watermarks, and page-level changes. - **Text or table extraction**: use pypdf for simple text and pdfplumber when layout or table boundaries matter. - **Scanned or image-only PDFs**: render pages and use OCR only when selectable text is absent. Say clearly when OCR quality is uncertain. Do not use a low-level page API to lay out a document that is naturally a component tree. Do not use a browser screenshot as a substitute for a real PDF unless the user explicitly wants an image-only result. ## Generate designed PDFs with pdfcn pdfcn is a shadcn-compatible registry of copy-paste PDF components and blocks. It is not a single `pdfcn` runtime package. The registry copies components into the host project and installs the renderer dependencies for the selected base. Use the package manager already used by the project. In a Bun worksp