| name | narrative-craft |
| description | Storytelling expertise for transforming graph-like product knowledge into compelling linear narratives. Covers storytelling frameworks (SCR, Hero's Journey, Problem-Solution), audience-framework selection, narrative arc design, and the So What test. Keywords: storytelling, narrative, story arc, framework, audience, SCR, hero journey, graph-to-linear. |
Narrative Craft
Expert knowledge for transforming interconnected product knowledge into compelling, audience-specific linear narratives. This is the core intellectual skill of the Product Storytelling Agent.
When to Use This Skill
Load this skill when:
- Designing a narrative strategy for a presentation
- Choosing a storytelling framework for a specific goal + audience combination
- Transforming graph-like knowledge (specs, briefs, research) into a linear story
- Analyzing what an audience cares about and how they make decisions
- Structuring the emotional and rational arc of a presentation
The Central Challenge: Graph → Linear Transformation
Product knowledge is inherently graph-like: entities connect to each other in multiple ways, information is non-linear, and any single fact relates to many others. But humans absorb stories linearly — one idea after another, in a deliberate order that builds understanding and conviction.
The storyteller's job is to find the strongest path through the knowledge graph for a given audience and goal. This is not summarization — it's curation and sequencing.
Three Linearization Strategies
1. Dependency Ordering
Start with what the audience must understand first, then layer dependent concepts. Use when the content has prerequisite knowledge.
- Best for: Technical reviews, architecture decisions, complex product explanations
- Pattern: Foundation → Building blocks → Emergent capability → Implication
2. Audience-Priority Ordering
Start with what the audience cares about most, then provide supporting context. Use when the audience's attention is limited.
- Best for: Executive briefings, funding asks, status updates
- Pattern: Bottom line → Evidence → Context → Next steps
3. Tension-Resolution Ordering
Start by creating a problem or question, build tension, then resolve it. Use when you need emotional engagement.
- Best for: Customer stories, vision pitches, change management
- Pattern: Hook → Context → Tension → Evidence → Resolution → Action
Storytelling Frameworks Quick Reference
SCR: Situation–Complication–Resolution
Best for: Decision asks, funding requests, prioritization discussions
| Beat | Purpose | Example |
|---|
| Situation | Establish shared reality | "We've grown 3x in 12 months" |
| Complication | Introduce the tension | "But our infrastructure can't scale past 10K concurrent users" |
| Resolution | Present the path forward | "A $200K platform investment enables 100K users by Q4" |
Slide mapping: 2–3 situation slides → 2–3 complication slides → 3–4 resolution slides → 1 CTA
SCQA: Situation–Complication–Question–Answer
Best for: Analytical presentations, risk assessments, strategic options
Extends SCR by explicitly framing the question the audience should be asking. The Question beat focuses attention before the Answer provides resolution.
Hero's Journey (Customer as Protagonist)
Best for: Customer empathy stories, product vision, market positioning
| Beat | Purpose |
|---|
| Ordinary World | Customer's current reality and frustrations |
| Call to Adventure | The trigger that demands change |
| Trials | Challenges faced during transformation |
| Transformation | How the product/solution changes their world |
| New Reality | The measurably better outcome |
Slide mapping: 1–2 ordinary world → 1 call → 2–3 trials → 2–3 transformation → 1–2 new reality → 1 CTA
Problem–Solution
Best for: Technical reviews, architecture decisions, feature justifications
Direct and efficient. Present the problem with evidence, then the solution with evidence. Works when the audience values logic over narrative flair.
Slide mapping: 1 hook → 2–3 problem definition → 1–2 impact → 3–4 solution → 1–2 evidence → 1 CTA
What Is / What Could Be
Best for: Vision decks, aspiration pitches, roadmap presentations
Alternates between current state and future possibility to create a pull toward the vision. Each contrast pair increases the gap between "now" and "possible."
Slide mapping: 1 hook → (what is → what could be) × 3–4 pairs → 1 bridge → 1 CTA
Before–During–After
Best for: Transformation stories, migration plans, improvement case studies
Shows the journey from old state through transition to new state. Effective when the audience needs to understand both the change and the path.
Audience-Framework Matrix
| Audience | Goal: Decision Ask | Goal: Vision/Roadmap | Goal: Technical Review | Goal: Customer Story |
|---|
| Executive | SCR | What Is/Could Be | SCR (problem-first) | Hero's Journey |
| Technical | Problem–Solution | What Is/Could Be | Problem–Solution | Before–During–After |
| Customer-facing | Hero's Journey | What Is/Could Be | Problem–Solution | Hero's Journey |
| Investor | SCR | What Is/Could Be | SCQA | Hero's Journey |
| Mixed | SCR (safest) | What Is/Could Be | SCQA | SCR |
When in doubt, SCR is the safest framework — it works for almost any professional audience.
Narrative Arc Design
Every great presentation follows an emotional arc. Map your slides to this intensity curve:
Engagement
▲
│ ╱╲
│ ╱ ╲ ╱╲
│ ╱╲ ╱ ╲ ╱ ╲
│ ╱ ╲╱ ╲ ╱ ╲
│ ╱ ╳ ╲___
│╱
└──────────────────────────▶ Slides
Hook Context Tension Evidence Resolution CTA
Emotional vs. Rational Beats
Emotional beats create engagement and urgency:
- Customer pain quotes
- Striking statistics ("40% of users abandon within 3 minutes")
- Vision statements
- Before/after contrasts
Rational beats build conviction and credibility:
- Data analysis
- Competitive comparison
- Technical architecture
- ROI calculations
Rule: Alternate between emotional and rational beats. Never have more than 2 consecutive beats of the same type. If you have 3 data slides in a row, insert a human story or provocative question between them.
The "So What?" Test
Every slide, every section, every bullet point must answer one question for the target audience: "So what?"
Apply the test:
- Read the content from the audience's perspective
- Ask: "Why should I care about this?"
- If you can't answer in one sentence, the content fails the test
- Fix it: either cut it, reframe it in terms of what the audience values, or add the missing "so what"
Common "So What?" failures:
- Feature descriptions without user impact → Add the benefit
- Metrics without context → Add the comparison or trend
- Process details the audience doesn't need → Cut or summarize
- History without relevance to the decision → Cut or compress
The Punch Test
Every slide headline must make someone look up from their phone. If a headline is a flat statement or a topic label, it fails.
Apply the Punch Test:
- Read the headline out loud
- Ask: "Would a busy executive glance up and want to know more?"
- If no — rewrite using one of these patterns:
- Provocative question: "What if we could 3x conversion without adding headcount?"
- Bold claim: "We left $2M on the table last quarter"
- Surprising stat: "73% of users quit before seeing our best feature"
- Tension statement: "Our biggest competitor just solved the problem we're ignoring"
Punch Test examples:
- ❌ "Q3 revenue was $4.2M, representing a 40% increase" → ✅ "Revenue exploded 40% — and we're just getting started"
- ❌ "Customer satisfaction improved across all segments" → ✅ "Customers went from frustrated to fanatical in 90 days"
- ❌ "Market Overview" → ✅ "The market just shifted — and we're positioned to win"
- ❌ "Product Roadmap Update" → ✅ "Three bets that will define our next 12 months"
Emotional Beat Mapping
Plan which emotion each slide evokes. A deck without an emotional arc is a document, not a story.
The emotional progression for a decision-ask deck:
curiosity → concern → hope → excitement → confidence → action
Map it to slides:
| Slide Position | Target Emotion | How to Evoke It |
|---|
| Opening hook | Curiosity | Surprising stat, provocative question, bold claim |
| Context | Recognition | "You already know this" — nod-along content |
| Tension | Concern / Urgency | Cost of inaction, competitive threat, missed opportunity |
| Solution reveal | Hope | "What if we could..." — possibility framing |
| Evidence | Excitement | Proof it works — data, case studies, prototypes |
| Plan | Confidence | Clear path forward — timeline, resources, milestones |
| CTA | Action | Specific ask — make it easy to say yes |
Rule: Label each slide in your proposal with its intended emotion. If two adjacent slides target the same emotion, one of them is redundant.
The Movie Trailer Method
Structure your deck like a movie trailer — tease the payoff, build tension, deliver the climax, end with action.
The pattern:
- Tease (slides 1–2): Open with the payoff — "What if we could 3x our conversion?" Show the destination before the journey.
- Build tension (slides 3–5): "But right now, we're losing $2M/year to..." Show what's broken, what's at stake, why inaction is dangerous.
- Climax (slides 6–8): "Here's the solution" — reveal the approach with evidence. This is the turning point.
- Resolution (slides 9–10): Show what the world looks like after. Paint the future.
- Action (final slide): "Here's exactly what we need from you, by when."
The trailer method works because it creates a desire gap — the audience sees where they could be before they see how to get there. This makes them lean forward through the tension section.
Language That Punches vs. Summarizes
Headlines and body text should create momentum, not report facts. The difference between a boring deck and a compelling one is often just the language.
Punchy language patterns:
- Action verbs over state verbs: "Revenue exploded" not "Revenue was high"
- Specifics over generalities: "23% in 6 months" not "significant improvement"
- Forward-looking over backward-looking: "and we're just getting started" not "which was a good result"
- Contrast for drama: "from 14 days to 3" not "reduced onboarding time"
- Short, punchy sentences: "That changes everything." not "This represents a significant shift in our approach."
More examples:
- ❌ "The team successfully completed the migration project" → ✅ "Migration complete. Zero downtime. 40% faster."
- ❌ "User engagement metrics showed improvement" → ✅ "Users went from 2 visits/month to daily — in 6 weeks"
- ❌ "We recommend investing in the platform" → ✅ "Invest $200K now or lose $2M next year"
Throughline
Source: research §A lines 44–45; ported in session
2026-05-04-5707a9ef. Promotes what was previously implicit in the
design checklist into a first-class strategist deliverable.
A throughline is a single declarative sentence that the entire
deck is arguing. Every slide either states the throughline, supports
it, or sets up the next thing that will support it. If a slide does
none of those, the slide is on a different deck.
The throughline is not the title. The title is the brand label for
the talk. The throughline is the argument the audience walks out
having heard.
Examples:
| Title (brand label) | Throughline (argument) |
|---|
| "Q3 Platform Update" | "Our infrastructure can't reach Q4 targets without a $200K platform investment, and the cost of waiting until Q4 is a $2M revenue cap." |
| "Customer Onboarding Story" | "Acme moved from a 14-day to a 3-day onboarding by replacing manual provisioning with our self-serve flow — and they're now expanding to a second BU." |
| "Architecture Review" | "Migrating to event-sourced storage halves p99 latency while opening three downstream product moves; the alternative (sharding the monolith) costs more and unlocks none of them." |
Throughline rules:
- One sentence. Two if the audience demands a setup-then-claim
structure, never more.
- Names the audience's current state AND the desired state OR action.
"X is true" is not a throughline; "X is true, therefore the
audience should do Y" is.
- Survives translation to a verbal one-liner. If you can't say it out
loud in eight seconds, simplify.
- Lives at the top of the proposal under
## Throughline (see
proposal.schema.md).
The Slide-Sorter Test (deck-level QA)
A deck's slide titles, read in order, should be a coherent executive
summary of the throughline. This is the simplest deck-level test
that exists, and it's deterministic.
The test:
- Extract every slide title in deck order.
- Read them aloud as a paragraph (skipping section dividers).
- Ask: "Without seeing any slide bodies, would a busy executive who
only read this paragraph follow the argument from current state →
tension → resolution → action?"
- If yes → the slide-sorter test passes.
- If no → at least one title is a topic label (rewrite per Punch Test)
or the slide ordering is wrong.
This is the deterministic complement to the AEI triad below. AEI
guarantees each slide carries an argument; the slide-sorter test
guarantees the arguments compose into the throughline.
The critic runs this test as part of slide-critique/SKILL.md's
verdict logic — see "Slide-Sorter Test (Deck-Level)" there.
Per-Slide AEI Structure
Source: research §C lines 67–73; ported in session
2026-05-04-5707a9ef. Closes the gap that one-message-per-slide
tells you to pick one thing but doesn't tell you what parts
the one thing must contain.
Every non-title, non-section-divider slide must carry three things:
| Part | What it is | Where it lives |
|---|
| Assertion | The slide's claim, as a declarative sentence | Slide title (action title, not topic label) |
| Evidence | The proof — chart, number, quote, diagram, example, image, process, comparison, or table | Slide body — visual evidence preferred over text |
| Implication | Why this matters for this audience | Subtitle / callout / explicit "Implication:" line — or, when space is tight, the speaker notes |
The strategist names the AEI triad in the per-slide outline of
proposal.md. The deck-builder records it in deck-spec.json under
the optional core_assertion / evidence / implication fields
(v2.1.0). The critic checks that every non-title slide has a
non-empty assertion that survives the Punch Test, that the evidence
type matches the assertion's claim shape (see
pptx-engine/references/chart-selection.md for chart-type-by-relationship),
and that the implication answers "so what?" for the audience.
AEI Examples
❌ Topic-driven (no AEI):
Title: "Q3 Customer Feedback"
Body: 5 bullets describing what customers said.
(Assertion: missing. Evidence: a list. Implication: missing.)
✅ AEI-shaped:
Assertion (title): "Customers trust the product but onboarding
friction is delaying expansion" — research lines 51–52.
Evidence (body): Two quotes (one frustration, one praise) +
retention curve showing the onboarding-week dropout.
Implication (callout): "Closing the onboarding gap unlocks
~$1.4M in pre-existing pipeline."
Evidence Type Selection
Choose evidence type from the allowed set in deck-spec.schema.json's
slide evidence.type enum (chart | metric | quote | table |
diagram | image | example | comparison | process). The
intake's proof_required field (when present) names what the
audience considers credible — match evidence type to that signal
when possible.
When evidence is chart, the chart relationship dictates the chart
type — see pptx-engine/references/chart-selection.md.
AEI Anti-Patterns
- Topic-supported-by-bullets — title is a label, body is a
list. Convert the list to an annotated visual.
- Triple-implication slide — three "so what" lines and no
evidence. Pick one implication; demote the others.
- Implication = restated assertion — the "so what" must add
something the assertion doesn't already carry (a consequence, a
required action, a cost of inaction).
Anti-Patterns: Stories That Kill Decks
"Information Vomit"
8 bullets per slide, each a full sentence. The presenter reads them aloud. The audience reads ahead and tunes out. Fix: One message per slide, max 4 bullets, max 15 words each.
"The Encyclopedia"
Comprehensive but boring. Every fact included, no editorial judgment about what matters. Fix: Cut ruthlessly. If it doesn't drive the decision, it's an appendix item.
"The Hedger"
"Results were somewhat positive" / "We believe this could potentially improve..." / "There may be an opportunity to..." Fix: Take a position. "Results crushed expectations" or "Results fell short — here's why and what we're doing."
"The Agenda Slide Opener"
Starting with "Today we'll cover..." kills momentum before the story begins. Fix: Open with a hook — a surprising stat, a provocative question, or a bold claim. Never open with logistics.
"The Data Dump"
Six charts on one slide. Each chart has 10 data points. No headline telling you what to see. Fix: One chart per slide. Headline states the conclusion. Simplify to 3–5 data points max.
Handling Multiple Valid Narratives
When the knowledge graph supports several equally compelling paths:
- Generate 2–3 alternatives (never more than 3 — choice overload)
- For each, provide: framework name, 2-sentence rationale, and slide title sequence
- Include your recommendation with reasoning
- Let the user choose — they know their audience better than you do
References
For detailed framework deep-dives:
