agentleFS
mblodemblode/agent-skills/skills/presentation-creator/SKILL.md

presentation-creator

Skill189 starsChanged 2 months agoMITView on GitHub

Builds decks with a story spine, house visual system, setting-specific density, and speaker notes. Use when asked to "create a presentation", "write a pitch deck", or "turn this doc into slides". Defaults to Marp; use an available presentation tool for editable PowerPoint.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read each step's reference when you reach it:

    line 46
  2. Very specific:

    Read web-deck.md only after the copy exists.

    line 59
  3. Distinctive:

    Every spine beat has at least one slide, and the ending matches the one written first

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Presentation Creator

Bold, minimal slide decks with a story underneath: spine to final QA.

  • IS: slide decks end to end: story spine, slide sequence, slide copy, visual system, speaker notes, investor pitch decks, and decks built as a web app; output as Marp markdown (default), Slidev or reveal.js markdown, or a Next.js deck app.
  • IS NOT: producing or editing the .pptx/.potx file itself (available presentation/PPTX skill or tool; hand it the finished outline, copy, and notes from this skill), charts inside a slide (external dataviz where installed), long-form prose, marketing copy, or a spoken talk script outside slides (ghostwriter), or product UI (ui-design).

Workflow

Track this checklist:

Presentation progress:
- [ ] Step 1: Gather context (audience, setting, venue, three messages, output format)
- [ ] Step 2: Write the story spine and the ending (references/story-structure.md)
- [ ] Step 3: Outline the slide sequence (references/outline-structure.md)
- [ ] Step 4: Write slide copy (references/writing-slides.md; pitch decks: references/pitch-decks.md instead)
- [ ] Step 5: Design colour, type, and layout (references/visual-design.md)
- [ ] Step 6: Write speaker notes (references/speaker-notes.md); skip for a deck sent without a presenter
- [ ] Step 7: Emit the deck in the chosen format (references/output-formats.md or references/web-deck.md)
- [ ] Step 8: QA pass, output the slide-by-slide review table

Step 1: Gather context

Establish these, asking only for what the brief does not answer:

  • Audience: internal (shared context, be direct) vs. external (build credibility, define terms).
  • Setting: live talk, recorded/async, or a pitch deck sent to investors and read without you.
  • Venue: a dark room or a large hall takes the dark system; a bright meeting room, daylight, or a deck that doubles as a handout takes a light palette. Ask when unknown, because it decides Step 5.
  • The three messages: what the audience must remember after the deck.
  • Output format: Marp markdown unless there is a reason otherwise. A web app when the deck should run a live demo or live at a URL. A .pptx when the user names the file or a house template, produced by the external pptx skill from this skill's outline, copy, and notes. Decide now, not at Step 7: the format sets the type and notes mechanics in Steps 5 and 6.

Route by setting:

  • Live talk, internal, or recorded deck → Steps 2-8 in order.
  • Investor pitch deck sent to be read → references/pitch-decks.md first. Its 10-slide framework replaces Step 3's outline, and its async copy rules replace writing-slides.md at Step 4 (do not load both: their density rules contradict). Step 2 still applies in compressed form; the spine is what stops a pitch reading as a feature list. Skip Step 6. The same company pitching live on a demo-day stage is a presented deck: use the standard path with the pitch framework as its outline.

Steps 2-7: Build the deck

Read each step's reference when you reach it:

StepReferenceCovers
2. Storyreferences/story-structure.mdThe spine template, writing the ending first, taking a position, stakes, the unstick move
3. Outlinereferences/outline-structure.mdNarrative flow, 12 slide types, section colours, outline output format
4. Writereferences/writing-slides.md, replaced by references/pitch-decks.md on the pitch pathHeadline patterns, body rules, copy per slide type, before/after examples
5. Designreferences/visual-design.mdTwo colour systems, contrast thresholds, fluid and fixed type scales, layout patterns, slide type to layout mapping
6. Notesreferences/speaker-notes.mdPer-slide note structure, delivery cues, notes by slide type
7. Emitreferences/output-formats.mdMarp syntax and export, Slidev and reveal.js equivalents, the pptx handoff, where notes live in each
7. Emit (web)references/web-deck.mdRoute-per-slide structure, navigation, layout primitives, motion, live demos
Changing this skillevals/evals.jsonBehavioural scenarios with assertions, plus should-trigger and near-miss routing prompts. Never loads during a user task

Read web-deck.md only after the copy exists. Primitives designed before the outline get shaped around slide 3 and fight every slide after it.

Step 8: QA pass (produces evidence)

Render the delivered format and inspect every slide. Record layout defects and measured contrast; a self-assigned glance score is editorial judgement. Keep the detailed table with the artifact when useful, and summarize material results in chat:

| # | Slide | 3-sec test | One message | Spine beat | Layout | Colour | Contrast |
|---|-------|-----------|-------------|------------|--------|--------|----------|
| 1 | Title | pass | pass | once upon a time | full statement | teal | 12.6:1 |
  • 3-sec test: parseable in three seconds at arm's length (Duarte's glance test). Cut copy on any failure until it passes. Pitch decks are read, not glanced: substitute "makes sense forwarded with no context".
  • One message: exactly one idea per slide; split slides carrying two.
  • Spine beat: which beat of the Step 2 spine this slide serves. A slide serving none is a fact you found interesting; cut it.
  • Layout: the layout should change when the slide's job changes. Flag a run of three or more identical layouts and keep it only when the section is deliberately a list.
  • Colour: the section accent, or the full-bleed palette, matching what Step 5 assigned.
  • Contrast: the smallest text on the slide against its background. 4.5:1 for body and captions, 3:1 only for text at 24px (18pt) or larger. Record the ratio, not "ok".

Deck-level checks below the table:

  • Every spine beat has at least one slide, and the ending matches the one written first
  • The deck states a position a reasonable person could disagree with
  • One colour system throughout: accents per section, or full-bleed palettes, never both
  • Recap slide has exactly one line per core section
  • Speaker notes sit where the output format reads them (Marp and Slidev: an HTML comment at the end of the slide; reveal.js: a Note: line; .pptx: the notes pane)
  • Pitch decks only: 10 slides plus an appendix at most, explicit ask slide (amount and use of funds), headlines pass the forwardable test

Fix observed defects and inspect affected slides again. If rendering is unavailable, label visual verification unrun.

Gotchas

  • Dark deck in a bright room: the default dark system relies on the room. Under daylight or a weak projector the black background goes grey and white body text washes out. Ask about the venue in Step 1; take the light "paper" palette or a white background when the answer is bright, and test on the projector, not the laptop.
  • Contrast checked at headline size only: a saturated full-bleed pair that reads at 100px fails at 20px caption size. Measure the smallest text on the slide from the actual hex values against the Step 8 thresholds.
  • Export "PPTX" from Marp or Slidev and call it done: both rasterise each slide into an image inside the .pptx. Text is not selectable or editable, so the deck the client wanted to edit is a stack of pictures. When editable PowerPoint is the deliverable, route to the pptx skill.
  • Notes and directives both live in HTML comments in Marp: <!-- _class: lead --> is a directive, <!-- Open with the outage story --> is a presenter note. A note that starts with a key: value line silently becomes a directive.
  • Fixed pixel type on a web deck: a deck sized for the presenter's laptop is a different deck on the projector and unreadable on the phone it gets forwarded to. Size in clamp(); Marp and .pptx decks are fixed canvases and take pt sizes instead.
  • Presented-deck density on a pitch deck sent by email: a 3-words-per-slide deck forwarded with no presenter is unreadable. Route to pitch-decks.md at Step 1, not after the deck is built. The inverse also fails: a 60-word slide on a demo-day stage.
  • Skipping the spine: jumping straight to slides produces a list of facts with no arc, then a rewrite once the missing narrative shows. Spine and ending first.
  • Speaker notes as a script: a verbatim script gets read aloud and sounds flat. Notes are prompts: key point, talk-track bullets, transition line.
  • Accents outside the section system: section colours are wayfinding; a random mid-section accent reads as a topic change that never happened. On a full-bleed deck the slide is the accent; the colour changes at the slide boundary, not inside it.
  • Paragraphs on slides: the audience reads instead of listening and the speaker becomes redundant. Cut until the 3-second test passes.

Related skills

  • External pptx skill (anthropics/skills) where installed: creating, editing, and QA of the .pptx file. This skill owns story, outline, copy, and notes; on a visual conflict inside a .pptx, this skill's colour system and type hierarchy set direction and the pptx skill's font, margin, and notes mechanics win.
  • External dataviz skill where installed: any chart or metric tile on a slide.
  • ghostwriter: marketing copy outside a deck, and long-form prose from the blog profile when the output is prose, not slides.
  • ui-design: visual systems for product UI and landing pages; presentation visual rules live in references/visual-design.md instead.

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.