Presentation Forge logo

Presentation Forge

Organization
studio-yebisu
presentation-forge

Build beautiful, self-contained HTML presentations (slide decks) with the bundled Presentation Forge engine: one HTML file per slide, bundled into a single portable index.html with presenter mode, progressive reveal, and swappable themes, plus built-in animations (five motion levels), charts, figures and icons. Use this whenever the user wants to create, design, write, or build a presentation, slide deck, talk, or "slides" as HTML or for the browser; wants a web-based or single-file shareable deck; wants to turn a topic, brief, outline, notes, or a document into slides (technical talks AND any other subject); or wants to recreate a brand/PowerPoint look as a reusable theme, from a .pptx, from image(s), or from a text description, optionally integrating a company logo. Produces HTML decks, not native PowerPoint files.

Overview

Publisherstudio-yebisu
Repositoryyaoyorozu-ai-toolbox-slides
Skill namepresentation-forge
Stars
10
Forks
2
Bundled files
36
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 36 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by studio-yebisu on GitHub. Read the source before you install it.

Installation

Install the Presentation Forge AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/studio-yebisu/yaoyorozu-ai-toolbox-slides/tree/main/examples/repositories/presentation-forge-showcase/source/upstream
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/studio-yebisu/yaoyorozu-ai-toolbox-slides.git /tmp/yaoyorozu-ai-toolbox-slides
mkdir -p .claude/skills
cp -r /tmp/yaoyorozu-ai-toolbox-slides/examples/repositories/presentation-forge-showcase/source/upstream .claude/skills/presentation-forge
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Presentation Forge in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Presentation Forge on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Presentation Forge is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Presentation Forge

Presentation Forge turns plain HTML into polished presentations. You write one HTML file per slide; a tiny Python build step (build.py, standard library only) bundles the slides, the engine, the chosen theme and all images into a single self-contained index.html that opens by double-click, emails cleanly, and works offline.

Everything needed ships inside this skill, so it works the same in Claude Code, in the Claude apps, and via the API - anywhere Claude has a filesystem and can run Python.

Three layers, always kept separate:

  • engine (template/engine/) - rendering logic: scaling, navigation, presenter mode, progressive reveal. Don't edit it to change content or look.
  • theme (template/themes/<name>/) - the look: colours, type, spacing, fonts, logos, backgrounds. Swap themes without touching slides.
  • content (slides/) - the slides, one file each, ordered by name.

This skill supports two workflows, chosen from what the user asks:

  1. Create a presentation - author a deck from a topic or brief.
  2. Import a theme - build a reusable theme from a .pptx, image(s), or a description, with optional logo integration. See reference/import-theme.md for the full procedure.

Where the engine lives

This skill bundles the deck scaffold in template/, a sibling of this SKILL.md: engine/ (with the animation toolkit, the components and the icons), themes/ (the dark obsidian and the light ink-blue), an empty slides/, assets/, build.py and deck.config.json. The scripts are in scripts/.

Resolve the skill's own directory first (it's wherever this SKILL.md was read from), then treat template/ and scripts/ as relative to it:

sh
SKILL_DIR="$(dirname "$(find . -name SKILL.md -path '*presentation-forge*' 2>/dev/null | head -1)")"
# or just use the directory you read this SKILL.md from

In Claude Code the skill folder is known directly; in the Claude apps / API the skill is unzipped into the working filesystem - in both cases template/ sits next to SKILL.md. Use python instead of python3 wherever that is what resolves on this machine (common on Windows).

Workflow 1 - Create a presentation

The user decides what the deck says. They describe their talk - topic, points, notes, a document - when they start the conversation. Build from that; don't quiz them about content. If the brief is thin, write a plausible first draft and flag what you invented so they can correct it.

Step 1 - Ask two questions, then nothing else

Before writing anything, ask these two together, in the user's language, in one message (with the AskUserQuestion tool when it is available, otherwise a short message listing the options). Skip a question the user already answered, and skip both if they said to go straight ahead - then use the defaults.

  1. How should we build it?
    • Plan first (default, recommended) - you propose the outline, they adjust it, then you write the whole deck.
    • All at once - you write the whole deck in one go.
    • Slide by slide - you write one slide, they review it, then the next.
  2. How much animation?
    • None - perfectly still.
    • Subtle - soft fades, nothing distracting.
    • Balanced (default) - polished entrances, counting figures, reveals.
    • Lively - more movement, ambient loops, decorative shapes.
    • Extra - the full show: 3D, springy pops, drawn icons, strong loops.

Step 2 - Create the deck

Pick a target folder (a new kebab-case folder named after the topic in the current working directory, unless the user named one) and create it empty:

sh
python3 "$SKILL_DIR/scripts/new_deck.py" "<target>" --title "<title>" \
    --lang <en|fr|…> --theme <obsidian|ink-blue|…> --motion <level>

The deck starts with no slides: everything in it comes from this user's brief. Never build inside the skill's template/.

Step 3 - Plan the deck from the brief

Fix the one core message, the audience and an arc that follows what the user wants to say (see reference/writing-decks.md). For each slide choose its shape from reference/layouts.md: a figure, a chart, a process, a comparison, bullets… The only fixed slides are the title and the close; an agenda, section dividers, a quote or a footer appear only when this content calls for them - not because a deck "usually" has them.

Step 4 - Write, following the chosen mode

Slides go in <target>/slides/, one <section class="slide"> per file, numbered 01-, 02-, … (see The slide authoring contract). Use the components in reference/components.md and the motion in reference/motion.md - matching the chosen level - rather than inventing CSS.

  • Plan first - show the outline: for each slide, its title (a full assertion), its shape, and one line on what it shows. Wait for the user's answer, apply their changes, then write every slide, build and check (Step 5).
  • All at once - write every slide, build and check.
  • Slide by slide - show a short outline once, then for each slide: write it, build, check it (--slides N), show the user that slide's image and wait for their feedback before the next one. Apply what they say to the current slide, and carry their preferences forward.

Step 5 - Build, then look at the result

sh
python3 build.py                                     # inside <target>
python3 "$SKILL_DIR/scripts/check_deck.py" "<target>"   # add --slides N or all

The build must print Built index.html - N slides with the count you expect, with no ! warnings. check_deck.py renders the deck in headless Chrome, lists layout problems (text past the slide's edge, content cut off, broken images) and saves a contact sheet of every slide plus an image of each problem slide. Fix every reported problem and look at the contact sheet before handing the deck over. In slide-by-slide mode, share the slide image with the user when your environment can show files; otherwise tell them to open index.html#N. If no Chrome / Chromium / Edge is installed, say the visual check was skipped - never install a browser.

Step 6 - Hand it over

Report the deck folder, the theme and motion level, anything you invented for a thin brief, and how to present: arrow keys / Space, p presenter mode, ? shortcuts. Mention they can fix wording right in the browser (Chrome or Edge, on the local file) - those edits are kept.

Changing an existing deck

The user may have edited text in the browser since the last build; those edits live in index.html. Before touching slides/, run python3 build.py --pull: it copies them back into the slide files, so you work on the latest text (every build does this too, so a rebuild never loses them). If it reports edits it could not merge, the edited deck was kept as index.unmerged-edits.html - bring those lines over by hand, then delete that file.

Write it well, from the start

Treat the writing as the product, not an afterthought. Per slide:

  • make the title a full assertion that states the point ("Caching cut p99 by 40%", not "Performance"); the body is only the evidence for it;
  • one idea per slide, few words, parallel and concrete bullets, no walls of text - prose belongs in the notes;
  • give almost every content slide speaker notes in <aside class="notes">: the spoken narration and delivery cues the slide does NOT show (not a copy of the slide text, not a word-for-word script);
  • match the language to the audience; flag specifics you invent for a thin brief.

For the full method (arc, assertion-evidence, tight on-slide text, and exactly what to put in speaker notes), read reference/writing-decks.md before writing. It is what makes the deck and its text excellent rather than merely correct.

Workflow 2 - Import a theme

Reproduce a brand's charter as a reusable theme under template/themes/<name>/ (or a deck's themes/<name>/): its palette, its real fonts (downloaded and embedded, not a system fallback), its logo, and its visual signature (layout: title placement, bands, rules, footer). The reference can be:

  • a saved style file (.pfstyle.json) from a previous session - the exact, one-step path: scripts/theme_bundle.py unpack rebuilds the whole theme (CSS, fonts, logo, backgrounds) byte for byte, with no reproduction needed;
  • a PowerPoint (.pptx) - scripts/pptx_theme.py extracts the palette, the fonts, the embedded media, and the master's layout geometry (title/body boxes, font sizes, background);
  • image(s) - scripts/image_colors.py samples the exact dominant colours; view the image for typography and layout;
  • a text description - brand words mapped to tokens;
  • optionally a company logo to integrate onto slides.

Real fonts are fetched with scripts/fetch_font.py (Google Fonts) when free, with a fallback to asking for the font files. Whenever you build a theme, export it with scripts/theme_bundle.py pack into a single <name>.pfstyle.json and give it to the user: handing back that one file in any future conversation recreates the exact same style, with no image or PowerPoint needed. The full step-by-step procedure (colour mapping, font fetching, logo integration, reproducing the layout in slides.css, and packing/unpacking the style file) is in reference/import-theme.md. Read it when this workflow triggers.

The slide authoring contract

Each file in slides/ is exactly one slide. Keep markup plain and lean on the theme's classes - that keeps slides consistent and themes swappable.

html
<section class="slide">
  <h2 class="title">One clear point per slide.</h2>
  <ul class="bullets">
    <li>One idea per line.</li>
  </ul>
  <aside class="notes">Speaker notes - shown only in presenter mode.</aside>
</section>

Slides are authored on a fixed 1920×1080 canvas; the engine scales it to any screen, so always design against that fixed size.

Slide variants

ClassUse
slidestandard content slide
slide slide--titleopening / hero slide
slide slide--sectionsection divider (dark background)
slide slide--concludeclosing slide

Content blocks (from the theme)

.eyebrow (kicker) · .display (largest heading) · h1/.title, h2/.subtitle, .lead · ul.bullets · .two-col (two-column grid) · .card (callout) · blockquote · pre > code (escape < > &) · .footer · .muted / .accent (colour helpers) · aside.notes (presenter-only notes).

Components (from the engine)

Figures (.stats), bar charts (.bars), gauges (.ring), processes (ol.steps), timelines (ol.timeline), comparison tables (table.compare), columns of cards (.grid), 84 line icons (<i data-icon="rocket">) and decorative shapes (.fx). They follow the active theme and only the ones a deck uses are bundled. Markup and limits: reference/components.md.

Motion

Entrances (data-anim="rise"), one-after-the-other lists (data-stagger), counting figures (data-count), ambient loops (data-loop), per-slide transitions (data-transition) - all scaled by the deck's level. What to use at each level: reference/motion.md.

Progressive reveal

Add class="fragment" to any element to reveal it step by step on click (add data-anim to choose the effect, or data-anim="highlight" to mark words instead). Each → reveals the next fragment, then advances to the next slide; the presenter view shows step 2/3.

Images

Put a deck's content images in assets/ and reference them relative to the deck root (<img src="assets/diagram.png">); build.py inlines them as base64. Keep theme assets (fonts, backgrounds, logos) inside the theme folder.

Configuration (deck.config.json)

scripts/new_deck.py writes it; change it any time and rebuild.

KeyDefaultPurpose
titlePresentationpage title
langendocument language (fr, en, …)
themeobsidianwhich themes/<name>/ folder to use
motionbalancednone · subtle · balanced · lively · extra
transitionfadenone · fade · slide · zoom · rise · blur · flip
width height1920 1080design canvas size
exit_hintEnglish stringtoast shown on entering full screen

An optional deck.css at the deck's root holds styles for that deck only (a one-off layout); it loads after the theme. Prefer theme classes and components.

Theming basics

A theme is a self-contained folder: tokens.css (colours, type scale, spacing, font-family names), fonts.css (@font-face), slides.css (block styling), plus fonts/, images/, logos/. To make a look, copy themes/ink-blue/ to themes/<name>/, edit tokens.css, and set "theme": "<name>". Every theme must define the same token names and style the same slide classes, so switching a theme never breaks a deck.

Reference

  • reference/writing-decks.md - how to write the presentation and its text well, and what to put in speaker notes. Load it for Workflow 1.
  • reference/layouts.md - the shapes a slide can take and when each fits. Load it to plan a deck.
  • reference/components.md - figures, charts, processes, timelines, tables, icons, decoration: markup and limits.
  • reference/motion.md - the animation toolkit and what to use at each motion level.
  • reference/import-theme.md - full theme-import procedure (pptx / image / description / logo). Load it for Workflow 2.
  • template/docs/writing-slides.md - the deep authoring guide (navigation, presenter mode, deeper theming). Load it when you need detail beyond the contract above.

Guardrails

  • One <section class="slide"> per file in slides/; the build warns otherwise.
  • Don't edit engine/ to change content or styling - that's the theme's job.
  • Don't build inside template/; always start a deck with scripts/new_deck.py.
  • Prefer the theme's classes, the components and the motion toolkit over inline styles or custom CSS, so themes stay swappable and the level stays a dial.
  • Never hand over a deck without building it and running check_deck.py (when a browser is available).
  • Never pip install a package into the system/global Python - it can clobber a version another project on the machine relies on. Every script here is stdlib-only except scripts/image_colors.py (needs Pillow), which already handles this: on import failure it installs Pillow into a dedicated venv at ~/.cache/presentation-forge/venv via scripts/pf_venv.py and re-execs itself there. If a future script needs another package, follow the same pattern (pf_venv.reexec_in_venv("<pypi-name>")) instead of a bare pip install.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Presentation Forge AI skill do?

Build beautiful, self-contained HTML presentations (slide decks) with the bundled Presentation Forge engine: one HTML file per slide, bundled into a single portable index.html with presenter mode, progressive reveal, and swappable themes, plus built-in animations (five motion levels), charts, figures and icons. Use this whenever the user wants to create, design, write, or build a presentation, slide deck, talk, or "slides" as HTML or for the browser; wants a web-based or single-file shareable deck; wants to turn a topic, brief, outline, notes, or a document into slides (technical talks AND...

Why use Presentation Forge on TypingMind?

Because you install it once and use it with any model. Presentation Forge is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Presentation Forge in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/studio-yebisu/yaoyorozu-ai-toolbox-slides/tree/main/examples/repositories/presentation-forge-showcase/source/upstream. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Presentation Forge?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Presentation Forge?

As many as you like. As long as a model supports skills, you can use Presentation Forge with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Presentation Forge AI skill free?

It is published on GitHub by studio-yebisu. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇