agentleFS
hfadhlullahhfadhlullah/kaizen/skills/media/kaizen-deck/SKILL.md

kaizen-deck

Skill6 starsChanged 45 days agoMITView on GitHub

Build a cinematic, minimalist storytelling pitch deck (A5 landscape .pptx) in the style of a senior agency — story first, problem, a single "what if" question, then the product reveal and feature scenes, ending in a demo CTA. Dark ink + gold, Playfair Display + Inter, full-bleed graded photos. Use when the user asks for a pitch deck / sales deck / product showcase deck "that tells a story", "cinematic", or a new deck for a product or audience in that style. Do NOT use for editing an existing .pptx, plain bullet decks, or data reports — use the pptx skill for those.

  • Installs packages

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the Anthropic pptx skill is available, load it once (pptxgenjs gotchas, validator).

    line 15
  2. Very specific:

    Run scripts/check.sh first on a new machine.

    line 16
  3. Very specific:

    Write plan.json (backgrounds + panels), run python3 prep_images.py plan.json → bg/.

    line 31
  4. Very specific:

    Fonts: Run scripts/install_fonts.py once per machine (static instances.

    line 32
  5. Distinctive:

    structured questions: ask in plain text with numbered options.

    line 81

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Kaizen deck

A deck that tells a story before it sells. Worked example: examples/build-example.js (every helper on a fictional product, no images needed).

Read before building:

  • references/story-arc.md — the acts, slide recipes, copy rules, layout facts.
  • references/rules.md — truth, images, fonts, files, QA. Every rule there cost a redo.

If the Anthropic pptx skill is available, load it once (pptxgenjs gotchas, validator); qa.sh works without it. Run scripts/check.sh first on a new machine.

Workflow

  1. Brainstorm first. Propose 2–3 story concepts + an outline (one line per slide) and ask: audience, language, data available, photos, the CTA. Build only after the outline is agreed.
  2. Deck folder: ~/.kaizen/decks/<slug>/ — everything a deck needs and produces lives there (never in the user's project folders). Shared dependencies live once in ~/.kaizen/decks/: node_modules (pptxgenjs, resolved upward by Node) and .venv (Pillow, defusedxml, lxml, fonttools, qrcode). First time on a machine: cd ~/.kaizen/decks && npm init -y && npm i pptxgenjs && python3 -m venv .venv && .venv/bin/pip install pillow defusedxml lxml fonttools "qrcode[pil]". Copy in scripts/kit.js, scripts/prep_images.py, scripts/qa.sh.
  3. Research the truth. Features from the product repos (shipped only — one sub-agent per repo when there are several, if your agent has sub-agents). Stats from official sources with citations.
  4. Photos. Find on Unsplash with a web-fetch tool (or ask the user for photos), download to img/, contact-sheet, choose. Write plan.json (backgrounds + panels), run python3 prep_images.py plan.json → bg/.
  5. Fonts. Run scripts/install_fonts.py once per machine (static instances; see rules.md).
  6. Write build-<name>.js with kit.js:

    const { makeDeck } = require("./kit");
    const d = makeDeck("Deck title", { logo: "logo.png", logoRatio: 199 / 880 }); // branding optional
    d.story("s01", { kick, title, sub, notes, credit, top });  // photo story beat
    d.pause(line, sub, notes);                                  // black silence slide
    d.question(line, notes);                                    // gold "what if" slide
    d.reveal(name, tagline, line, notes);                       // product name reveal
    d.scene("f03", { kick, title, features: [[name, desc]], notes, credit, shot, shotRatio }); // shot = UI screenshot, framed right
    d.cards(kick, title, [[label, name, desc] x3], notes, footnote);
    d.cta("f16", { title, sub, facts, notes, credit, contact, site, qr }); // qr: PNG path (make with python qrcode), else a [QR] box
    d.pres.writeFile({ fileName: "Deck.pptx" }).then(console.log);

    Branding options on makeDeck(title, brand): INK, CREAM, MUTED, GOLD, GOLD_BG, cardBg (hex, no #), SERIF, SANS (font names), logo (light-on-dark PNG path; none = no logo), logoRatio (height/width). Defaults are the neutral ink + gold look with no logo. For anything the helpers do not cover (triptych, recap timeline, stat rows) use d.slide(), d.kicker(), d.headline(), d.body(); scripts/kit.js shows how they work.

  7. Motion (when asked for animation or transitions): copy in scripts/motion.py and run python3 motion.py Deck.pptx Deck-motion.pptx, then QA the result. See "Brand theme and motion" below.
  8. QA: ./qa.sh build-<name>.js, look at every printed sheet (or send them to the user), fix, repeat. Clean qa/ after.
  9. Report: file path, the story in one line per slide, every placeholder left, every claim the user must verify, and what was deliberately left out.

Brand theme and motion

Follow the user's brand when they name one. Asked to match a landing page or site, read its CSS and markup first, and use its colors, type, shapes and exact words instead of ink and gold. Pass them through the makeDeck branding options, or build with d.slide(). Write new copy only for what the source lacks (a researched problem, a contact slide), and say which lines are yours. Sales decks end on contact: name, company, site, and a QR code. Do not put decks in the project repo.

Transitions and animation: motion.py adds fade and push transitions and automatic entrance fades. Name shapes a<N>-... with objectName so the same N appears together; unnamed shapes appear one by one.

Product footage is MP4, never GIF. GIFs may not play in presentation apps. Render clips from real footage with the hyperframes skill, then embed with slide.addMedia({ type: "video", path, cover: "data:image/png;base64,...", x, y, w, h }). cover must be base64. motion.py makes the first video on each slide autoplay, muted and looping. Renders and QA sheets cannot show motion, so report it as unplayed until someone runs the slideshow.

Agent notes (Claude Code, Codex, Gemini CLI, OpenCode, Cursor…)

Steps name capabilities, not tools. Where your agent lacks one:

  • web fetch (finding stock photos, sources): ask the user for photos/links instead.
  • viewing images (checking contact sheets / frames): send the image paths to the user and ask them to look.
  • sub-agents (reading several repos at once): read them one after another.
  • structured questions: ask in plain text with numbered options.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.