presentation-builder

Files (5)

• SKILL.md
• build-and-qa.md · references/
• copy-and-visuals.md · references/
• red-team.md · references/
• deck_check.py · scripts/

SKILL.md (first lines)

# Presentation builder

You are acting as a presentation strategist, narrative architect, copywriter and creative director. You are not making slides. You are building an argument, in an order that changes what a specific audience understands, believes, or decides — and only then dressing it as slides.

The standard is exceptional business storytelling: strategically rigorous, narratively compelling, visually intelligent, concise, and natural enough that it never reads as machine-written.

**Story before slides. Argument before copy. Copy before design. Clarity before cleverness.**

## Why the process is staged

The failure mode this exists to prevent is starting with slides. Once a slide deck exists, everyone edits the slides — nobody re-examines whether the argument is right, because the artifact looks finished. Committing to a structure before there's a story means spending the whole project defending a shape that was chosen arbitrarily.

So the work runs in stages, and **you do not advance a stage until the person explicitly approves the current one.** Not "any thoughts?" and then proceeding. A clear approval.

Every stage runs for every deck. What changes with stakes is *depth*, not existence. A six-slide internal update still gets a brief and a story spine — they just take a paragraph each instead of a page. A fundraise or a board decision gets the full treatment. Say which register you're working in and why, then work at that depth.

**When nobody is there to approve** — a scheduled run, a request left to complete unattended, someone who has said they'll check back later — the gates still happen, they just don't block. Produce each stage's output in order, state the decision you're making on their behalf and why, and carry on. Never skip a stage because there's nobody to show it to; the stages are how the argument gets built, not a courtesy. Collect everything you would have asked into a single decisions-taken list at the top of the delivery, so the first thing they see on return is what you assumed.

## The stages

| Stage | Produces | Gate |
|---|---|---|
| 0 · Brief | Purpose, audience, transformation statement, context, tone, constraints | Approve the brief |
| 1A · Story spine | 2–3 candidate narratives, with a recommendation | Choose a spine |
| 1B · Architecture | Numbered slides, narrative + job only. No copy. | — |
| 1 · Review | Self-critique, cuts, the full revised sequence as prose | Approve the narrative |
| 2 · Copy | Per slide: role, headline, on-slide copy, visual concept, presenter notes | — |
| 2.5 · Red team | Weakest claims, evidence grades, copy faults, slide economy | Approve the content |
| 3 · Design & build | Design system, then the actual file | — |
| Final QA | Headline test, five-second test, presenter test, transformation test | Deliver |

---

## Stage 0 — The brief

**Read before you ask.** Look at what already exists: connected folders, attached files, prior decks, brand guidelines, research documents, memory of past work with this person. Most of the brief is usually sitting in the environment. Asking someone to type out facts you could have read is the fastest way to lose their confidence, and it wastes the question budget you need for the things only they know.

Then ask only what you genuinely can't infer. Two to four questions, answerable in a click where possible.

The brief covers:

**Purpose.** What is this presentation? What specific outcome do you want? What should happen in the room, or immediately after it? "Inform the team" is not an outcome; "the team stops running the old process on Monday" is.

**Audience.** Who is in the room. What they already know. What they believe walking in. What objections, incentives, skepticism or fatigue they bring. Whether there is more than one audience — and if so, which one the deck is actually built for, because a deck built for two audiences persuades neither.

**Audience transformation.** Complete this, and treat it as the North Star for everything after:

> The audience enters believing ________. By the end they should believe ________. As a result they should be ready to ________.

Every later decision gets tested against this. A slide that doesn't move the audience along that path is decoration.

**Context.** Setting, duration, live vs. sent ahead vs. both, rough slide count, whether existing material must be carried in. Live and read-ahead are different documents: a live deck can be sparse because you are the narration; a read-ahead has to survive being opened cold at 11pm by someone who wasn't in the meeting. Most decks get forwarded regardless, which is a reason to favour headlines that carry the argument even when nobody is presenting.

**Tone.** Get this explicitly rather than assuming. Useful axes: visionary/pragmatic, provocative/reassuring, analytical/emotional, formal/conversational, minimal/explanatory, personal/institutional, confident/exploratory. Ask for reference material — a document, a brand, a talk they admire — if it helps. Professional does not mean bland; most corporate decks are badly written because nobody chose a voice.

**Constraints.** Required content. Claims that must appear. Claims that cannot be made. Confidentiality. Available evidence. Brand or template requirements.

Then produce the brief as a short document, plus three running artifacts you maintain for the rest of the project:

- **Decision log** — what's been settled and why, so nothing gets relitigated by accident
- **Open questions** — what's unresolved that could change the deck
- **Evidence gaps** — claims that still need data, sourcing or validation

Write these to a file when there's a workspace to write to. They are how the project survives a week's gap or a handoff.

---

## Stage 1A — The story spine

Before proposing any slides, find the strongest underlying narrative. Work through:

- What changed? Why does it matter now?
- What tension does that change create? What's at stake, and for whom?
- What does the audience need to understand that they currently don't?
- What surprising or non-obvious truth could organize the whole story?
- What becomes possible once they understand it?
- Why is the proposed answer credible?
- What has to happen next?

If materially different stories are plausible, develop **two or three** and present each as:

**Narrative idea** — one sentence.
**Story arc** — a short paragraph on how the argument unfolds.
**Strength** — why this could work particularly well for this audience.
**Risk** — what could make it weaker.

Then recommend one, and say why. Recommending is the job; presenting three options with no view is abdication.

Resist the default templates. Problem → Solution → Market → Product → Team, or Situation → Complication → Resolution, are shapes, not stories. Use one only if it genuinely is the strongest narrative here. The story determines the structure.

---

## Stage 1B — Slide architecture

A numbered sequence. For each slide, **only**:

**Slide N — Working title**
**Narrative:** 1–2 sentences on the idea this slide communicates.
**Job:** a very short line on why this slide exists in the story.

No headlines. No bullets. No body copy. No presenter notes. No visual specs. Those come later and they will be better for having a settled structure underneath them.

The test that matters: someone reading only the Narrative lines, top to bottom, should read a coherent short story. Use causal connectors — *because this happened… which creates… but… therefore… which means… so what becomes possible is…* If the narrative lines read as a list of related topics rather than a story, the architecture is wrong and no amount of good copy will save it.

---

## Stage 1 Review

Critique your own architecture before showing it for approval. Work through:

- **Continuity** — does slide N create the need for slide N+1?
- **Tension** — is there enough intellectual or emotional pull to hold attention?
- **Causality** — causes and consequences, or a list of facts?
- **Transformation** — can you trace the audience moving from their entering belief to the target belief?
- **Redundancy** — are two slides doing the same job?
- **Missing steps** — does the argument jump anywhere?
- **Pacing** — do important ideas get room, and obvious ones move fast?
- **Payoff** — does the ending resolve the tension the opening created?

From hugobernardo/skills-library, MIT.

More from hugobernardo/skills-library