extract-template · diff
git:20260921.3e35aa5 to git:20261012.0411448
45 added, 37 removed. Audit A to A.
---
name: extract-template
- description: "Decide what an uploaded file is — a deck, a Word document, a PDF document, or artwork — and hand it to the template-extraction guide that matches. Use when asked to extract a template from a deck or document, turn a PDF into a Word template, save an image style as a reusable template, apply a company template to Markdown, or save a file's visual language as a reusable template."
+ description: "Extract a reusable presentation, document, or illustration template from source files or an HTML deck URL. Use when asked to extract a deck or Word template, turn a PDF into a Word template, apply a company template to Markdown, or save an image style for reuse."
---
# Extract a reusable template from a file
- Look at what the file holds, decide what it is, follow that guide to the end.
- Run the commands below from the directory containing this guide.
+ Choose the intended template, select its extraction guide, then follow that guide
+ through inspection, extraction, validation and saving.
- Render a `.pptx` or `.ppt` first. An image is already viewable.
+ ## 1. Choose the template type
- ```bash
- okou presentation screenshot --input <source.pptx|source.ppt> --out shots
- ```
+ Use the user's requested template type. For example, a report can supply the
+ visual style for a presentation template. When the user leaves the type open,
+ identify what the source contains:
- Render a PDF or Word document at its own page proportions before judging layout:
+ | Template type | Source content |
+ | --- | --- |
+ | Presentation | Slides organized around individual ideas, with display text and visual compositions |
+ | Document | A report, letter, resume, invoice or other document with reading order and document structure |
+ | Illustration | Artwork whose drawing style will be reused |
- ```bash
- node scripts/render-document.mjs --input <source.pdf|source.docx> --out source-render
- ```
+ Classify PDFs and page images by their content. A slide screenshot belongs to
+ presentation; a scanned document belongs to document. Sparse text or a short page
+ count alone does not make a document a slide deck. Several artwork references in
+ one drawing style form one illustration template.
- Use a new output directory for each render. Keep the original as the source;
- use the returned PDF and page images for inspection.
+ ## 2. Select the source and guide
- | The file is | Go to |
- |---|---|
- | a deck — `.pptx`, `.ppt`, an image deck, or a `.pdf` whose pages are slides | `presentation/SKILL.md` |
- | a Word document — `.docx` | `docx/SKILL.md` |
- | a PDF document — pages of prose, not slides | `pdf/SKILL.md` |
- | artwork — `.png`, `.jpg`, or several images of one drawing style | `illustration/SKILL.md` |
+ Keep the original file. When both an authoring file and its export are available,
+ prefer the authoring file: `.pptx` over its PDF, `.docx` over its PDF.
- A page carrying one idea in display type is a slide. A page carrying running
- prose under a repeated header, footer, or page number is a document. A picture with no
- reading order is artwork.
+ | Chosen template | Available source | Guide |
+ | --- | --- | --- |
+ | Presentation | HTML file, HTML directory or deck URL; `.pptx`, `.ppt`, PDF or ordered slide images | [Presentation extraction](presentation/SKILL.md) |
+ | Document | Word file | [Word extraction](docx/SKILL.md) |
+ | Document | PDF | [PDF extraction](pdf/SKILL.md) |
+ | Illustration | One or more artwork images | [Illustration extraction](illustration/SKILL.md) |
- Re-save a `.doc`, or a `.docx` written by WPS, as `.docx` from Word first.
+ Prepare another format only when the selected guide requires it:
- Four rules override the table:
+ - Scanned document images: assemble an ordered PDF for the PDF guide.
+ - Cross-format reuse: export slides to PDF for a Word-style template, or a Word
+ report to PDF for a presentation template. The requested template type stays
+ the same.
- - A deliverable the user names beats the table. "Turn these slides into a Word
- template" is `docx/`; "save this report's look as a presentation template" is
- `presentation/`.
- - Extract from the authoring file when it sits beside an export of itself: `.docx`
- over its PDF, `.pptx` over its PDF.
- - One or two pages at a paper size is never slides, however sparse. A resume and
- an invoice carry as little text as a slide and are still documents.
- - Several images at once are one style to `illustration/`, not several
- templates. A single image of a page — a scanned document, a screenshot of
- slides — is that page's kind, not artwork.
+ Keep conversions alongside the original source.
- `cd` into the branch directory; every path inside a branch guide is relative to
- it. Both document guides start with
- [`document-reuse.md`](document-reuse.md) to choose the reuse scope and package.
+ ## 3. Follow the selected guide
+
+ `cd` into the selected guide's directory; its commands and paths are relative to
+ that directory.
+
+ The guide handles source rendering when needed. Rendering produces page images
+ for inspecting layout, typography, colours and image crops; those captures become
+ the visual reference for extraction. Keep the original for structure and style
+ metadata. Reuse existing source captures; no need to render them again.
+
+ Document guides use [document-reuse.md](document-reuse.md) to choose between slot
+ filling, structure and expression reuse, and style extraction. The presentation
+ guide separately identifies extraction files and the generation files bundled
+ into the saved template.