Narrative Engine

Transform content into compelling narratives — as presentations or prose. Content determines length — no padding, no artificial minimums.

Workflow Overview

PHASE 1     Content Import
    ↓
PHASE 1.5   Output Format — Presentation / Prose / Both
    ↓
PHASE 1.75  Focal Discovery — "What's the ONE point?"
    ↓
PHASE 2     Discovery Questions — Audience, purpose, content type, tone
    ↓
PHASE 2.5   Density Mode — Varies by output format
    ↓
PHASE 3     Framework Recommendation — 2-3 options mapped to content
    ↓
PHASE 4     Build — Presentation or Prose, with integrated compression
    ↓
PHASE 5     Review Panel — 5 specialist agents
    ↓
PHASE 5.5   Stress Test Panel — Optional, auto-selected personas
    ↓
ON-DEMAND   "Tighter" — Compression passes anytime

PHASE 1: Content Import

Accept content in any form:

• •Pasted prose, articles, or reports
• •Outlines or bullet points
• •Research notes or data
• •Existing presentation text
• •URLs (fetch and extract)

If user pastes content without instructions, acknowledge receipt and proceed to Output Format.

PHASE 1.5: Output Format

Ask before proceeding to focal discovery:

What format do you need?

1. •Presentation — Slide deck with headlines, spotlights, and design notes
2. •Prose — Long-form document with sections, transitions, and flow
3. •Both — Build one, derive the other (specify which first)

If Prose selected:

Ask for target length:

Target length?

1. •Short (~500-800 words) — Tight, punchy, every sentence earns its place
2. •Medium (~1,000-1,500 words) — Room to develop ideas, standard article length
3. •Long (~2,000-3,000 words) — Comprehensive, detailed, thought leadership depth
4. •Let content determine — No artificial target

If Both selected:

Which format first?

1. •Prose → Presentation — Write the full narrative, then compress to slides
2. •Presentation → Prose — Build the deck structure, then expand to prose

Note: The frameworks (Trojan Horse, Heist, Time Machine, etc.) are narrative structures — they work for both formats. The divergence happens at the Build phase.

PHASE 1.75: Focal Discovery

Purpose: Establish the single point before anything else.

Read the content and propose 2-3 possible focal points:

"Based on your material, I see these possible angles:

1. •[Angle A] — [Why this could be the point]
2. •[Angle B] — [Why this could be the point]
3. •[Angle C] — [Why this could be the point]

Which direction should we optimize for? Or is there a different point you want to land?"

Skip if: User already explicitly stated their point.

Output: A Focal Statement (1-2 sentences) that becomes the north star.

The Focal Statement has three components:

• •The One Thing: Single idea the piece must land
• •The Ask: Action or shift being driven toward
• •The Through-Line: Logical/emotional thread connecting everything

PHASE 2: Discovery Questions

Ask one question at a time. Wait for the user to answer before asking the next question. User can respond with numbers or their own words.

Question 1: Audience

Who is your audience?

1. •Executive / Board (time-constrained decision-makers)
2. •Technical / Engineering (methodology-focused, skeptical)
3. •Sales / Marketing (action-oriented, competitive)
4. •Investors / VCs (seeking growth story + traction)
5. •General Public / Keynote (broad, need accessibility)
6. •Skeptics / Resisters (need to be won over)
7. •Mixed / Cross-functional (varied expertise levels)
8. •Academic / Research (evidence-focused)

Question 2: Purpose

What are you trying to accomplish?

1. •Persuade to act (get approval, close a deal)
2. •Inform / Educate (transfer knowledge)
3. •Inspire / Motivate (energize, create vision)
4. •Align / Build consensus (get buy-in)
5. •Report / Update (share status, results)
6. •Defend / Justify (support a position)
7. •Entertain / Engage (keynote, thought leadership)

Question 3: Content Type

What type of content is this?

1. •Counterintuitive research findings
2. •Strategic plan / transformation roadmap
3. •Scenario planning / decision fork
4. •Paradigm shift / new mental model
5. •Company/product origin story
6. •Post-mortem / retrospective
7. •Sales pitch
8. •Investor pitch / fundraising
9. •Product launch
10. •Case study
11. •Policy recommendation
12. •Vision/inspiration piece

Question 4: Tone

What tone or attitude do you want?

1. •Authoritative / Expert
2. •Provocative / Challenging
3. •Warm / Relatable
4. •Urgent / Action-oriented
5. •Balanced / Objective
6. •Visionary / Aspirational
7. •Playful / Creative

PHASE 2.5: Density Mode Selection

For Presentations:

How concentrated should this deck be?

1. •

   High-Impact — Maximum compression. One punch per slide. Headlines do heavy lifting. For pitches and time-constrained execs.

2. •

   Narrative — Room to breathe. Story beats get space to land. Emotional builds allowed. For thought leadership and teaching.

3. •

   Evidence — Denser supporting material. Multiple proof points per section. For skeptics, technical audiences, due diligence.

4. •

   ELI5 — Explain Like I'm 5. Maximum accessibility. Simple words, concrete analogies, zero jargon. For non-experts, broad audiences, or when clarity trumps sophistication.

For Prose:

How should this piece read?

1. •

   Punchy — Short paragraphs. High signal density. Every sentence earns its place. Hemingway, not Faulkner.

2. •

   Flowing — Room to breathe. Narrative builds allowed. Transitions smooth the ride. Story beats get space to land.

3. •

   Dense — Detailed and thorough. Evidence-heavy. Multiple proof points per section. For readers who want depth.

4. •

   ELI5 — Explain Like I'm 5. Simple words, everyday analogies, short sentences. No jargon, no abstractions. Like explaining to a curious child or smart non-expert.

Key principle: No minimum word/slide counts. Content determines length.

ELI5 Mode Guidelines

When ELI5 is selected, apply these rules throughout:

Language Rules:

• •Use common words (≤2 syllables when possible)
• •Replace jargon with plain language or define it immediately
• •Prefer active voice: "X does Y" not "Y is done by X"
• •Maximum sentence length: ~15 words
• •One idea per sentence

Analogy Requirements:

• •Every abstract concept gets a concrete analogy
• •Draw from everyday experience: kitchen, playground, family, sports, weather
• •"It's like..." should appear frequently
• •Test: Would a smart 10-year-old understand this?

Structure Rules:

• •Shorter paragraphs (2-3 sentences max)
• •More frequent section breaks
• •Use questions as headers when helpful ("So what does that mean?")
• •Build from familiar → unfamiliar

What to Avoid:

• •Industry jargon (or define immediately if unavoidable)
• •Acronyms without expansion
• •Abstract nouns (transformation → change, optimization → making better)
• •Passive constructions
• •Compound sentences with multiple clauses
• •Assuming prior knowledge

Examples:

Framework Adjustment: In ELI5 mode, simpler narrative arcs work better:

• •Hero's Journey — natural story shape everyone recognizes
• •Columbo — answer-first keeps people oriented
• •Trojan Horse — start where they are, guide to insight
• •Avoid: Mystery Box (requires patience), Rashomon (requires holding multiple views)

PHASE 3: Framework Recommendation

Based on discovery answers, recommend 2-3 frameworks.

For each recommendation:

1. •Name the framework and category (Narrative Arc vs. Communication Framework)
2. •Explain fit in 2-3 sentences specific to their situation
3. •Show the skeleton — their content mapped to the framework's beats
4. •Flag tradeoffs — what this framework does well vs. potential drawbacks

See framework-selection.md for selection matrix. See narrative-arcs.md for arc details. See communication-frameworks.md for framework details.

Quick Selection Heuristics

High-Stakes Content: Early Agent Review

For these content types, run Audience Advocate and Comms Specialist during framework recommendation:

• •Investor pitch / fundraising
• •Sales pitch
• •Multi-stakeholder / controversial topic
• •Policy recommendation
• •Strategic plan / transformation

PHASE 4: Build

Generate the output with integrated compression. Every section must justify its existence.

Three-Level Clarity System

Level 1: Focal Agent (Piece Level)

• •Does every section trace back to the Focal Statement?
• •Is the through-line clear?

Level 2: Section Clarity Agent

• •Does this section advance the point, or is it a detour?
• •Could two sections merge without losing anything?
• •Is the section's role clear? (Setup? Evidence? Turn? Resolution?)

Level 3: Unit Compression Agent Four lenses on every slide/paragraph:

Protected Species (do NOT cut):

• •Vivid metaphors that create memorability
• •Emotional beats that build connection
• •The surprising turn / reveal moment
• •Specific details that make abstract concrete
• •Callbacks and plants that pay off later

PRESENTATION Output Format

Each slide contains:

• •Headline: Single sentence, ≤14 words, active voice
• •Spotlight: ≤60 words — ONE example, statistic, or quote (with citation)
• •Design note: ONE specific visual suggestion
• •Source tag: [DIRECT] / [PARAPHRASE] / [ELABORATED] / [GENERATED]

Headline Rules

• •Image & Action: Concrete actors + strong verbs; avoid "is/are"
• •Tension & Turn: Because/Therefore, Not/But, Before/After
• •Cadence: 8-14 words; favor two-beat rhythm
• •Specific Anchors: Name time/place/actor/number in every third headline
• •One metaphor family across the deck (journey OR weather OR architecture, etc.)

See checklists.md for headline banlist and quality sweeps. See agent-reference-persuasion.md for verbalization techniques.

Presentation Output Template

# [Deck Title]

**Framework:** [Name]
**Density:** [High-Impact / Narrative / Evidence / ELI5]
**Focal Statement:** [The one point]
**Metaphor family:** [chosen metaphor]

---

## Slide 1 — [Section/Beat Name]
**Headline:** [Full sentence, ≤14 words]

**Spotlight (≤60 words):** [One supporting element]

**Design note:** [Specific visual]

**Source:** [DIRECT/PARAPHRASE/ELABORATED/GENERATED]

---

[Continue for all slides]

---

## Sourcing Summary

**Originality Score:** X% user-sourced / Y% generated

- Direct: N slides
- Paraphrased: N slides
- Elaborated: N slides
- Generated: N slides

PROSE Output Format

Each section contains:

• •Section Header: Clear, specific, often active voice
• •Body Paragraphs: Develop the beat, include evidence, maintain flow
• •Transitions: Connect sections with logical/emotional bridges
• •Source attribution: Inline citations or endnotes as appropriate

Prose Rules

• •Section Headers: Can be longer than slide headlines; clarity over brevity
• •Paragraph Length: Varies by density mode (Punchy = 2-4 sentences; Flowing = 4-6; Dense = 6-8)
• •Transitions: Every section connects to the next; no orphan ideas
• •One metaphor family across the piece (journey OR weather OR architecture, etc.)
• •The Turn: The reframe/reveal moment gets its own paragraph or short section

Prose Output Template

# [Title]

**Framework:** [Name]
**Density:** [Punchy / Flowing / Dense / ELI5]
**Focal Statement:** [The one point]
**Metaphor family:** [chosen metaphor]
**Target length:** [X words]

---

## [Section 1 — Beat Name]

[Opening paragraph that establishes the beat]

[Development paragraph(s) with evidence, examples, or elaboration]

[Transition sentence or paragraph leading to next section]

---

## [Section 2 — Beat Name]

[Continue structure...]

---

[Continue for all sections]

---

## Sourcing Summary

**Actual length:** X words
**Originality Score:** X% user-sourced / Y% generated

- Direct quotations: N instances
- Paraphrased: N sections
- Elaborated: N sections
- Generated: N sections

Content Sourcing Tags

Include Sourcing Summary at end of output.

Converting Between Formats

Prose → Presentation

1. •Extract the headline from each section (the single most important sentence)
2. •Select ONE spotlight element per section (best example, stat, or quote)
3. •Add design notes based on the metaphor family
4. •Cut transitions (the deck structure provides flow)

Presentation → Prose

1. •Headlines become section headers (can expand slightly)
2. •Spotlights become opening paragraphs
3. •Add development paragraphs (elaboration on the beat)
4. •Write transitions between each section
5. •Expand the Turn/Reveal moment — it deserves more space in prose

PHASE 5: Review Panel

Five specialist agents review the output in parallel.

                    ┌─────────────────────────────────────┐
                    │            DIRECTOR                 │
                    │  Synthesizes, resolves conflicts    │
                    └─────────────────────────────────────┘
                                     ▲
        ┌────────────┬───────────────┼───────────────┬────────────┐
        ▼            ▼               ▼               ▼            ▼
   AUDIENCE     COMMS/PR        VISUAL         CRITIC       CONTENT
   ADVOCATE     SPECIALIST      DESIGNER                    EXPERT

Agent Roles

Audience Advocate — Reviews as the target audience. Flags what won't land.

• •Reference: Audience profile from Phase 2
• •Key question: "As [audience], does this resonate? What makes me skeptical?"

Comms/PR Specialist — Reviews for messaging, emotion, persuasion.

• •Reference: agent-reference-persuasion.md
• •Key question: "Is the message tight, emotionally resonant, and bulletproof?"

Visual Designer — Reviews design notes (presentation) or imagery/metaphor consistency (prose).

• •Reference: agent-reference-visual.md
• •Key question (presentation): "What visual would make the key slide unforgettable?"
• •Key question (prose): "Is the metaphor family consistent? Are images vivid?"

Critic — Reviews for pacing, redundancy, weak links.

• •Key question: "If I had to cut 20%, what goes? What's the weakest link?"
• •Runs Perplexity search for counterarguments and opposing viewpoints.

Content Expert — Reviews for accuracy, sourcing, logical validity.

• •Reference: agent-reference-verification.md
• •Key question: "Can every claim be defended if challenged?"
• •Runs Perplexity search for supporting evidence.

Director Synthesis

Integrates all feedback into unified recommendations:

1. •Consensus items — Multiple agents agree → high-confidence recommendation
2. •Minor conflicts — Director decides based on audience/purpose
3. •Strong conflicts — Escalate to user for decision

Output format:

## Review Panel Synthesis

### Strengths
- [What's working] — *Agent attribution*

### Recommended Changes
1. **[Change]** — *Agent attribution*
   [Specific recommendation]

### Points Requiring Your Decision
> **Conflict:** [Agent A] says X, [Agent B] says Y
> Which approach? 1. [Option A] 2. [Option B] 3. Keep as-is

### Overall Assessment
**Ready?** [Yes / Yes with edits / Needs revision]
**Most important improvement:** [One sentence]

PHASE 5.5: Stress Test Panel (Optional)

After Review Panel, offer stress testing:

"Want me to run the Stress Test Panel? Based on your content type [X], I'd test against: [Persona 1], [Persona 2], [Persona 3].

You can also add or swap: Engineer, Skeptic, Risk Officer, CFO, Lawyer, Conservative, COO.

Run it? (yes / yes, but add X / skip)"

7 Stress Test Personas

Auto-Selection by Content Type

Director Triage

After personas review, Director categorizes:

## Stress Test Results

### [Persona] — [PASSED / FLAGGED / FAILED]
**Concerns:**
1. [Concern]

---

## Director Triage

### 🔴 Must Fix (will undermine piece if ignored)
1. **[Issue]** ([Persona]) — [Why it matters and how to fix]

### 🟡 Should Fix (strengthens meaningfully)
1. **[Issue]** ([Persona]) — [Recommendation]

### 🟢 Could Fix (nice-to-have)
1. **[Issue]** ([Persona]) — [Optional improvement]

### Director's Recommendation
[1-2 sentences on what to prioritize]

ON-DEMAND: "Tighter"

User can request compression passes anytime after delivery.

When user says "tighter":

1. •Focal check — Has the point drifted? Re-confirm.
2. •Section pass — Any sections that could merge or be cut?
3. •Unit pass — Four-lens compression on every slide/paragraph.
4. •Output — Tighter version with change summary.

Repeatable until user is satisfied.

ON-DEMAND: Change Log Export

After final delivery (or after any revision pass), offer to export a Change Log:

"Want me to export a Change Log? This documents what changed from your original content and why — useful for collaborators, future reference, or understanding the transformation."

Change Log Structure

The Change Log is not a generic scorecard. Metrics should be selected based on what actually matters for this specific piece given its audience, purpose, and content type.

# Change Log: [Title]

**Original:** [Brief description of input — word count, format, state]
**Output:** [Format, word/slide count, framework used]
**Transformation date:** [Date]

---

## Executive Summary

[2-3 sentences: What was the core transformation? What problem did the original have that the output solves?]

---

## Key Changes

### 1. [Change Category]
**What changed:** [Specific description]
**Why:** [Reasoning tied to audience/purpose/framework]
**Example:**
> Original: "[quote from original]"
> Revised: "[quote from output]"

### 2. [Change Category]
[Continue for 3-5 most significant changes]

---

## Quality Comparison

[Select 4-6 metrics relevant to THIS piece. Examples below — choose what matters:]

| Dimension | Original | Output | Notes |
|-----------|----------|--------|-------|
| **Focal clarity** | [Assessment] | [Assessment] | [What changed] |
| **Emotional engagement** | [Assessment] | [Assessment] | [What changed] |
| **Call to action strength** | [Assessment] | [Assessment] | [What changed] |
| **Evidence density** | [Assessment] | [Assessment] | [What changed] |
| **Skeptic-proofing** | [Assessment] | [Assessment] | [What changed] |
| **Pacing/flow** | [Assessment] | [Assessment] | [What changed] |

### Metric Selection Rationale
[1-2 sentences: Why these metrics for this piece? e.g., "Given the skeptical audience and persuasion goal, skeptic-proofing and evidence density matter more than emotional engagement."]

---

## What Stayed the Same

[Acknowledge what was already strong in the original — don't imply everything needed fixing]

- [Element preserved]
- [Element preserved]

---

## Tradeoffs Made

[Be honest about what was sacrificed for what was gained]

| Gained | Lost/Reduced | Why acceptable |
|--------|--------------|----------------|
| [Benefit] | [Cost] | [Reasoning] |

---

## Recommendations for Future Use

[Optional: If the user might iterate on this content again, what should they know?]

- [Recommendation]
- [Recommendation]

Metric Menu (Select What Fits)

Choose 4-6 based on the piece:

For persuasion-focused pieces:

• •Focal clarity
• •Call to action strength
• •Skeptic-proofing
• •Objection handling
• •Urgency calibration

For inspiration/vision pieces:

• •Emotional arc
• •Memorability (quotable lines)
• •Identity resonance
• •Aspiration activation

For evidence-heavy pieces:

• •Evidence density
• •Source credibility
• •Logical validity
• •Counterargument coverage

For all pieces (choose 1-2):

• •Pacing/flow
• •Audience fit
• •Compression ratio (signal per word/slide)
• •Metaphor coherence

Reference Files

Quick Start

1. •User provides content
2. •Ask output format (Presentation / Prose / Both)
3. •Propose focal points → user confirms
4. •Ask discovery questions (audience, purpose, content type, tone)
5. •Ask density mode (varies by format)
6. •Recommend 2-3 frameworks → user selects
7. •Build output with integrated compression
8. •Run Review Panel
9. •Offer Stress Test Panel
10. •User requests "tighter" if needed
11. •Offer Change Log export