Skip to content

Transform content into narrative-driven slide decks using proven storytelling and communication frameworks. A Claude Code skill.

Notifications You must be signed in to change notification settings

nraford7/Narrative-Engine

Repository files navigation

Narrative Engine

A Claude Code skill that transforms any content into compelling narratives — as presentations or prose — using proven storytelling frameworks.

Most presentations fail before they begin. Not because the content is wrong, but because the structure is. Narrative Engine matches your material to the right storytelling framework, auto-derives a distinct voice and emotional arc for your specific audience, then builds the output with a 6-agent review panel that catches what you'd miss.


What It Does

Paste an article, outline, research notes, or existing deck. Choose presentation or prose. Answer discovery questions. Get back:

  1. Full-sweep framework scoring across all 10 arcs + 7 communication frameworks, with a Dark Horse option
  2. A Build Brief translating your audience, voice, emotional arc, and strategies into concrete writing instructions
  3. A complete output (slides or prose) with auto-derived voice, emotional pacing, and opening/closing strategies
  4. A 6-agent review including an Originality Agent that catches generic AI patterns
  5. Sourcing transparency showing what came from your content vs. what was generated

How It Works

Phase 1: Discovery

Output format?               → Presentation / Prose / Both
Focal point?                 → Choose from 2-3 proposed angles
Who is your audience?        → Executive / Technical / Investors / Skeptics / Mixed...
What are you trying to do?   → Persuade / Inform / Inspire / Align / Report...
What type of content?        → Research / Strategy / Case study / Pitch / Vision...
What tone?                   → Authoritative / Provocative / Warm / Urgent / Balanced...
Density mode?                → High-Impact / Narrative / Evidence / ELI5

Phase 2: Framework Matching

The skill draws from 17 proven frameworks across two categories:

10 Narrative Arcs (engagement-optimized)

Arc Structure Best For
The Prestige Pledge → Turn → Prestige Counterintuitive findings
Mystery Box Clues → Red herrings → Click Research with unexpected conclusions
The Heist Goal → Obstacles → Crew → Execution Strategy & transformation
Time Machine Future-back → Present fork → Path Vision & scenario planning
Trojan Horse Relatable → Escalate → Reframe Paradigm shifts, skeptical audiences
Hero's Journey Call → Trials → Ordeal → Return Origin stories, change management
Columbo Outcome first → Reconstruction Post-mortems, root cause analysis
Game of the Scene Pattern → Name it → Heighten 3x Hidden dynamics, cross-domain insight
Rashomon Multi-view → Missing axis → Synthesis Controversial topics, stakeholder alignment
Freytag's Five-Act Exposition → Climax → Resolution Complex emotional narratives

7 Communication Frameworks (efficiency-optimized)

Framework Structure Best For
Minto Pyramid Answer → MECE supports → Evidence Executive updates, board decks
SCQA Situation → Complication → Question → Answer Opening hooks
AIDA Attention → Interest → Desire → Action Sales, fundraising
PAS Problem → Agitation → Solution Change management
Raskin Sales Deck Change → Stakes → Vision → Features → Proof B2B sales
Duarte Resonate What is ↔ What could be Keynotes, vision
SUCCESs Simple, Unexpected, Concrete, Credible, Emotional, Stories Quality checklist

Phase 3: Build Brief + Subagent Build

A Build Brief translates all discovery answers into concrete writing instructions:

  • Voice auto-derived from audience + tone (7 distinct profiles)
  • Audience profile with headline style, evidence preferences, CTA approach
  • Emotional arc calibrated to audience tolerance
  • Opening/closing strategies from the rhetorical strategy library
  • Killer line targets for memorability

The build then generates presentation slides or prose sections, each with source tags ([DIRECT] / [PARAPHRASE] / [ELABORATED] / [GENERATED]).

The build runs as a subagent via the Task tool — offloading the heaviest generation work from the main conversation and reducing context pressure.

Phase 4: Review Panel (6 Parallel Subagents)

Six specialist agents review your output in parallel:

                    ┌─────────────────────────────────────┐
                    │            DIRECTOR                 │
                    │  Synthesizes, resolves conflicts,   │
                    │  presents unified recommendations   │
                    └─────────────────────────────────────┘
                                     ▲
        ┌────────┬─────────┬─────────┼─────────┬─────────┬────────┐
        ▼        ▼         ▼         ▼         ▼         ▼
   AUDIENCE  COMMS/PR   VISUAL    CRITIC   CONTENT  ORIGINALITY
   ADVOCATE  SPECIALIST DESIGNER           EXPERT   AGENT
Agent Key Question
Audience Advocate "As [audience], does this land?"
Comms Specialist "Is this tight and bulletproof?"
Visual Designer "What visual makes this unforgettable?"
Critic "What's the weakest link?"
Content Expert "Can every claim be defended?"
Originality Agent "Would this feel different from a generic AI output?"

The Director synthesizes feedback, resolves conflicts, and surfaces decisions you need to make.

All six agents run as parallel subagents via the Task tool, returning findings simultaneously for the Director to synthesize in the main conversation.


Output Example

# Regional Governance: The Invisible Infrastructure

**Framework:** Time Machine
**Slides:** 14
**Metaphor family:** Nervous system / pulse / sensing

---

## Slide 1 — Opening Hook
**Headline:** It is March 2028, and the Mayor knows the water crisis will hit before anyone else does.

**Spotlight:** Three weeks before demand spikes, the morning brief flags the pattern.
Desalination capacity, population movement, industrial permits—the signals converged overnight.

**Design note:** Split screen—calm office on left, converging trend lines on right,
date stamp "March 2028" prominent.

**Source:** [GENERATED]

---

## Sourcing Summary

**Originality Score:** 43% user-sourced / 57% generated

- Direct from source: 3 slides
- Paraphrased: 2 slides
- Elaborated: 3 slides
- Generated: 6 slides

Benefits

Without Narrative Engine With Narrative Engine
Generic bullet points Single-point headlines with power verbs
Unclear structure Framework matched to audience + purpose via full sweep
Same voice every time 7 auto-derived voice profiles matched to audience + tone
No emotional design Emotional arcs calibrated to audience tolerance
Self-review blind spots 6-agent review including Originality Agent
Unknown AI additions Source tags show exactly what was generated
One-size-fits-all Build Brief ensures every choice propagates through output

Installation

Copy the Narrative-Engine folder to your Claude Code skills directory:

~/.claude/skills/Narrative-Engine/

Or clone this repository:

git clone https://github.com/nraford7/Narrative-Engine.git ~/.claude/skills/Narrative-Engine

Usage

Invoke with /Narrative-Engine in Claude Code, then:

  1. Paste your content (article, notes, outline, or existing deck)
  2. Choose output format (Presentation / Prose / Both)
  3. Confirm focal point from 2-3 proposed angles
  4. Answer discovery questions (audience, purpose, content type, tone)
  5. Select density mode (High-Impact / Narrative / Evidence / ELI5)
  6. Choose from 2-3 recommended frameworks (scored via full sweep)
  7. Review the Build Brief, then receive complete output + 6-agent review

Files

File Purpose
SKILL.md Main skill definition and workflow
narrative-arcs.md Beat-by-beat structures for all 10 arcs
communication-frameworks.md Detailed framework descriptions
framework-selection.md Full sweep protocol and selection matrices
audience-profiles.md Deep audience profiles with writing instructions
voice-profiles.md 7 voice profiles with auto-derive mapping
emotional-arcs.md Framework emotional textures + audience calibration
opening-closing-strategies.md Opening/closing strategy libraries + pairing matrix
checklists.md All quality checklists (headlines, CTAs, originality, killer line)
agent-reference-persuasion.md Comms agent frameworks (Cialdini, SUCCESs, Ogilvy)
agent-reference-visual.md Visual agent frameworks (Tufte, Duarte, metaphors)
agent-reference-verification.md Content agent frameworks (IFCN, SIFT, fallacies)
prompts/ Subagent prompt templates (builder, reviewer, stress-tester)
examples/ Full workflow examples (climate keynote, post-mortem, remote work)

Architecture

Phases 1–3.5 (discovery) run interactively in the main conversation. Phases 4+ dispatch subagents:

  • Build (Phase 4): Single subagent reads the Build Brief from /tmp/, generates output, self-reviews for originality
  • Review (Phase 5): 6 parallel subagents, each with a specialist lens and reference file
  • Stress Test (Phase 5.5): 3 parallel subagents with auto-selected personas

Framework Selection Quick Reference

If your audience is... And your goal is... Consider...
Executive / Board Any Pyramid or Columbo (answer-first)
Skeptics Persuade Trojan Horse + PAS
Investors Inspire Hero's Journey + Cinderella
Mixed / Cross-functional Align Rashomon or Heist
General / Keynote Entertain Prestige or Time Machine
If your content has... Consider...
A genuine surprise Prestige or Mystery Box
Multiple stakeholder views Rashomon
A transformation story Hero's Journey
Future vision Time Machine
Root cause analysis Columbo

License

MIT


Credits

Built on frameworks from:

  • Barbara Minto (Pyramid Principle)
  • Nancy Duarte (Resonate, Slide:ology)
  • Chip & Dan Heath (Made to Stick)
  • Robert Cialdini (Influence)
  • Andy Raskin (Greatest Sales Deck)
  • Edward Tufte (Data Visualization)
  • Joseph Campbell (Hero's Journey)
  • Christopher Nolan / J.J. Abrams (Narrative structures)

About

Transform content into narrative-driven slide decks using proven storytelling and communication frameworks. A Claude Code skill.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors