Skip to main content
skill CX Codex

storytelling-narrative — Claude Code skill: what it does

The storytelling-narrative skill shapes business material into a clear narrative, from slide decks to reports. Use it when you have the content but need a stronger through-line.

Apache-2.0 ★ 173 updated 2026-10-24

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add sammcj/agentic-coding --skill storytelling-narrative -a claude-code
Open on GitHub

What storytelling-narrative does

The skill helps turn existing knowledge, drafts, or data into a narrative with a central idea, tension, a turning point, and an outcome. It works with business formats such as pitches, blog posts, slide decks, and reports. It draws the story from the material rather than making one up.

The process starts by clarifying the audience, goal, format, and state of the material. It then examines the draft or data and works out the story's central line. Frameworks such as Pixar, SCQA, and the Pyramid Principle are used as ways to assess a narrative, not as forms to fill in. When several storylines seem plausible, the skill asks the user to choose.

For draft analysis, it refers to references/critique.md; format-specific notes are in references/formats.md. Install it for Claude Code with npx skills add sammcj/agentic-coding --skill storytelling-narrative -a claude-code. The listed skill file path is ~/.claude/skills/storytelling-narrative/SKILL.md. Expect questions when your choice would change the direction of the story.

What it helps with

  • find the central narrative for a slide deck
  • rework an article draft so its turning point is clear
  • shape report data into a coherent story
  • clarify what decision a pitch should lead to

Example prompts for the agent

  • Help me shape a pitch from these notes; first ask who will hear it and what decision it should lead to.
  • Review this article draft and identify its main storyline, turning point, and repeated ideas.
  • Look at the presentation data and suggest several possible narrative through-lines.

How to install storytelling-narrative

1Skills CLI (Vercel)

npx skills add sammcj/agentic-coding --skill storytelling-narrative -a claude-code

2Manually: put the files in place

~/.claude/skills/storytelling-narrative/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/storytelling-narrative/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about storytelling-narrative

How do I install storytelling-narrative in Claude Code?

Run npx skills add sammcj/agentic-coding --skill storytelling-narrative -a claude-code. The listed skill file path is ~/.claude/skills/storytelling-narrative/SKILL.md.

What does storytelling-narrative do?

It helps find and shape a central storyline in business material, including decks, articles, pitches, and reports. It can also help refine an existing draft.

Does storytelling-narrative require a fixed framework?

No. It treats approaches such as Pixar and SCQA as lenses for examining a story, rather than structures the material must follow.

Source text

SKILL.md · Apache-2.0
---
name: storytelling-narrative
description: "Structure, refine, or draft narrative content or storytelling for business communication such as slide decks, blog posts, pitches, reports. Use when the user wants help building a narrative arc, find the through-line, sequence beats or refine existing content for flow and impact."
---

# Storytelling & Narrative

Help the user build, refine, or restructure a narrative for a business artefact: slide deck, blog post, pitch, exec memo, article, or report. The skill is format-agnostic at its core. Format-specific notes live in `references/formats.md`.

The user is a domain expert. They usually know the material deeply or have rough touchpoints in mind. What they're missing is structure, flow, and a clear through-line. Your job is to help them surface what's already there and shape it, not to invent narrative from nothing or layer marketing language on top.

## Operating principles

**Lenses, not templates.** Storytelling frameworks (Pixar, ABT, Three-Act, StoryBrand, Hero's Journey, Golden Circle, SCQA, Pyramid Principle, Dykes Arc, Duarte Sparkline) are different surface arrangements of the same underlying primitives: status quo, tension, stakes, turn, resolution. Treat named frameworks as lenses to check a story against, not templates to fill in. Pick a lens when it fits the audience-goal-material combination - never force material into a framework that doesn't match.

**Narrative comes from the material, not a script.** Most business storytelling failures aren't structural - they're the result of not engaging with the material long enough to find what's actually there. Spend real time in Phase 2 (engaging the material) before reaching for any framework.

**Anti-slop posture.** No marketing adjectives, no hot takes, no cliche verbiage or buzzwords. Concrete nouns over abstract ones. Specific people, specific moments, specific numbers. If a sentence could appear in any company's blog post, rewrite it.

**Pause when there's a real fork; proceed when there isn't.** Use multi-choice questions (the `AskUserQuestion` tool) when the user genuinely needs to steer - multiple plausible spines, multiple framework lenses, multiple orderings, output-mode choice. Don't pause on small calls or things you can infer with confidence. The user has explicitly said: increase autonomy when you have understanding, ask when you don't.

## Narrative primitives

Every narrative - regardless of framework - is built from these. Find them in the material before structuring anything.

- **Status quo / "what is"** - the world the audience already lives in. Familiar enough that they nod.
- **Tension** - what's wrong, missing, or about to break. The reason the story exists.
- **Stakes** - what happens if nothing changes. Without stakes, the audience has no reason to care.
- **Turn** - the moment, decision, evidence, or insight that changes things. The most often-missing primitive in weak business narratives.
- **Resolution** - the new state, the action, the call. What the audience now does, believes, or decides.
- **Hook** - the entry point that opens the loop. Must land within the first 30-60 seconds (or first paragraph). Six common types in `references/hooks-and-moments.md`.
- **Through-line** - a single sentence the rest of the piece serves: "from X to Y because of Z". If you can't write the through-line, the narrative isn't ready.

The shape is always: **old understanding → tension → new understanding → new action**. Different frameworks just arrange this differently.

## Workflow

This is a phased flow, not a rigid sequence. Skip phases the user has already done; double back when new information surfaces. The pause points below are where multi-choice user input genuinely steers the work.

### Phase 1: Get oriented

Don't start structuring until you understand audience, goal, material state, and format. This isn't a "quick diagnostic" - these answers shape everything downstream. Ask substantively, but with focus.

Things you need to know:
- **Audience.** Who is this for? What do they already know? What do they care about? What position or fatigue are they bringing in?
- **Goal.** What should the audience think, feel, or do after consuming this? A specific decision, a belief shift, an action, an emotional response?
- **Material state.** Raw knowledge in their head? Rough touchpoints sketched out? An existing draft to refine? An existing deck/data to reshape?
- **Format and constraints.** Slide deck, blog post, pitch, memo, article, report? Length? Time-to-deliver? Live presentation or read-alone?
- **Tone.** Visionary / pragmatic / friendly / urgent / credible / sceptical / confident.

Use `AskUserQuestion` for things with discrete plausible answers (tone, audience type, format, material state). Use open prose questions when you need substance (audience-specific concerns, goal nuance).

If the user volunteered most of this in their initial prompt, don't re-ask - confirm what you understood and ask only what's missing.

### Phase 2: Engage with the material

This is the most undervalued phase. Skip it and you'll produce a structure that doesn't fit what the user actually has to say.

Branch by material state:

**Raw knowledge / rough touchpoints** - ask the user to talk through what they have. Don't summarise back yet - keep extracting. Probe for narrative substrate:
- "What changed? What made you start paying attention to this?"
- "What was the moment you realised X?"
- "Who's affected? What do they actually do differently?"
- "What's the part most people get wrong?"
- "What number or specific example surprised you?"

You're looking for moments, not abstractions. Concrete things - a meeting, a number, a customer, a failure, a near-miss - are the raw material narratives are built from.

**Existing draft** - read carefully. Find the implicit arc (or note its absence). Mark:
- Where the hook is (or where it should be)
- Whether there's a turn - and where
- What's the spine the rest serves
- What's repeated, what's buried, what's missing

Read `references/critique.md` for the diagnostic checklist.

**Existing data / deck** - look for the transformation already present. The story is usually there waiting to be uncovered: a before-after, a problem-solution, a counterintuitive finding, a shift in trend. Resist the urge to invent a story; surface the one that's there.

Output of this phase: a list of raw narrative material - concrete moments, facts, characters, tensions, changes - not yet ordered.

### Phase 3: Find the spine

Distill the raw material to:
- A single-sentence **through-line**: "from X to Y because of Z"
- A candidate **hook**
- The **turn** - what shifts
- The **stakes** - why it matters if nothing changes
- The **resolution** - new state, new action

If you can find more than one plausible spine, that's a fork - present 2-3 options to the user via `AskUserQuestion` and let them pick. Different spines lead to genuinely different narratives; this is one of the highest-impact decision points.

If there's only one spine that fits the material, propose it directly with rationale. Don't manufacture options for the sake of asking.

### Phase 4: Choose a structure (lens)

Match a framework lens to the audience-goal-material combination. See `references/narrative-frameworks.md` for the full catalogue and selection guidance. Quick orientation:

- **ABT (And/But/Therefore)** - short pieces: exec updates, elevator pitches, paragraph openers, slide subtitles.
- **Three-Act** - general default for blog posts, talks, longer writing.
- **Pixar (Once upon a time...)** - change stories, adoption stories, "how X became Y".
- **Golden Circle (Why/How/What)** - vision, purpose, mission, manifesto.
- **StoryBrand SB7** - customer-facing marketing where the customer is the hero.
- **Hero's Journey (condensed)** - founder origin, brand history, personal narrative.
- **Duarte Sparkline (what is / what could be)** - persuasive keynotes, change comms.
- **Dykes Data Storytelling Arc** - data-driven presentations and insight reports.
- **SCQA (Situation/Complication/Question/Answer)** - exec briefings, consulting memos.
- **Pyramid Principle (Minto)** - top-down written reports for executives who scan.

Lenses are not exclusive - a piece can use SCQA at the document level and ABT at the paragraph level. Present 2-3 lens options when more than one would work; explain the trade-offs briefly; let the user pick.

### Phase 5: Order the beats

Arrange the material into the chosen structure. Three things to attend to:

- **Sequence for emotional and logical flow.** The audience should always know why they're reading the next paragraph. Each beat earns the next.
- **Cut ruthlessly.** Most business narratives carry 30-50% material that doesn't serve the through-line. If a beat doesn't move the audience from where they were to where they need to be, it goes. Mark cuts as "park for later" rather than deleting outright - the user may reuse them elsewhere.
- **Place the turn.** The turn is usually mid-to-late, not upfront. Resist "executive summary" instinct that puts the conclusion first - that kills the narrative tension. Exception: read-alone formats for time-poor execs (memos, reports) often do need bottom-line-up-front; in that case, give a one-sentence headline then tell the story.

Show the user the proposed ordering. Pause via `AskUserQuestion` only if there are multiple genuinely plausible orderings - otherwise propose and let them react.

### Phase 6: Decide output mode

Before drafting anything, ask the user what level of help they want. Use `AskUserQuestion` with options like:

- **Structure only** - they take it from here and write the prose themselves.
- **Hook + key moments** - draft the opener and the most load-bearing beats; they fill the rest.
- **Full draft** - draft prose for every beat in their voice.
- **Refine existing draft to new structure** - rework what they already wrote against the new spine.

This is non-negotiable as a pause point. Don't assume the level of involvement the user wants.

### Phase 7: Draft (if requested)

When drafting, hold to:
- **Concrete over abstract.** Specific nouns, specific numbers, specific people. "The auth team rewrote the session middleware in two weeks" beats "The team transformed our authentication".
- **Verbs do the work.** Strong verbs over verb+adjective combos.
- **Vary sentence length.** Mix short with long. Don't write five sentences of the same length and structure in a row.
- **The user's voice, not yours.** Match the tone they signalled in Phase 1 and the rhythm of any existing material they shared.
- **Anti-slop check.** No marketing adjectives, no AI tells, no "Additionally" / "Furthermore" / "It's worth noting" sentence openers, no closing summaries that say "In conclusion, this approach...".
- **Show stakes and consequences.** Don't gloss over the tension - that's where the narrative energy lives.

For format-specific drafting concerns (slide one-idea-per-slide rule, blog opener strength, pitch tightness), read `references/formats.md`.

### Phase 8: Self-review

Before returning the final output, run a critique pass against `references/critique.md`. Common failure patterns:
- Hook is generic or buried
- Turn is missing or under-emphasised
- Stakes implied but not stated
- Through-line drifts mid-piece
- Slop language crept in during drafting

Fix these before delivering. If you find structural issues, surface them rather than silently rewriting - the user may have specific intent behind a choice that looks weak.

## When to use which reference

- `references/narrative-frameworks.md` - when picking a lens, or when the user wants to commit to a specific framework. Selection guidance, per-framework failure modes, and detail on the lesser-known lenses.
- `references/hooks-and-moments.md` - when crafting a hook or sharpening a turn. Six hook types, the turn test, placement principles.
- `references/formats.md` - when format-specific concerns matter (slide deck rules, blog opener, pitch arc length, memo bottom-line-up-front).
- `references/critique.md` - when reviewing existing content, or running self-review on drafted output.

## Multi-choice question patterns

The user has explicitly asked you to use `AskUserQuestion` where it earns its place. Heuristics for when to pause:

| Pause for | Don't pause for |
|---|---|
| Tone (visionary / pragmatic / urgent / credible) when not signalled | Word-level choices |
| Audience type when ambiguous | Tiny edits |
| Multiple plausible spines (genuine fork) | Obvious cuts |
| Framework lens with 2-3 viable fits | Single-fit lens |
| Ordering when multiple orderings are valid | Mechanical sequencing |
| Hook variant when 2-3 are equally plausible | When one hook clearly fits |

**Always pause for output mode** (structure-only / hook + key moments / full draft / refine existing draft) before drafting. Don't infer what level of help the user wants.

Ask 1-3 questions at a time, not 5+. Let the user steer; don't quiz them.

## Gotchas

- **Don't fall back to the same handful of frameworks every time.** Pixar, Three-Act, Golden Circle, StoryBrand, Hero's Journey, ABT are widely known, but they're a subset. SCQA, Pyramid Principle, Dykes Data Storytelling Arc, and Duarte's Sparkline are often the better fit for exec, consulting, data-driven, or report-style work. Pick by audience-goal-material fit, not by familiarity.
- **This skill is narrative-first and format-agnostic.** Charts, dashboards, and visual design are not in scope. If the piece is primarily data-driven visuals, hand off to the `storytelling-with-data` skill; otherwise stay on the narrative.
- **Don't skip Phase 2.** Engaging with the material is what makes the rest work. Frameworks applied to thin material produce slop.
- **Don't manufacture options to ask the user.** If there's one obvious answer, propose it. Multi-choice questions are for genuine forks, not for performance of consultation.
- **Don't lecture the user about narrative theory.** They want help with their content, not a lesson. Apply the lens silently; surface only what they need to make decisions.
- **Don't add closing "In summary..." paragraphs to drafted content.** That's a slop tell. The narrative should land naturally on its resolution.
- **The turn is the most under-built primitive in business writing.** Most weak drafts have a status quo, stakes, and a vague resolution but no clear shift. Watch for this when reviewing existing content.

© sammcj · Apache-2.0 · sammcj/agentic-coding

Similar skills

All skills for Writing and docs →

From the same set73

The whole set →
hst-voice hst-voice — a skill that trains Claude Code to write in Hunter S. Thompson's authentic voice: not a polished imitation, but the real technique with rough structure, wrong endings, and refusal to state a thesis. CX Codex +1 ★ 173repo stars iterative-refinement iterative-refinement structures measurable optimization loops against pass/fail checks, trims context consumption, and delegates subtasks. CX Codex +1 ★ 173repo stars shell-scripting shell-scripting — a Claude Code skill for writing defensive bash scripts with ShellCheck compliance and straightforward patterns. Guides you through reliable script structure, variable safety, and when to abandon shell for Python or Go. CX Codex +1 ★ 173repo stars home-assistant Connects Claude Code to a live Home Assistant setup to inspect entities, debug automation traces, and generate valid YAML configurations. CX Codex +1 ★ 173repo stars apply-mantel-styles apply-mantel-styles — a skill with guidelines for Mantel's colour palette, components, and layout. Use it to create diagrams or interfaces that match the brand identity. CX Codex +1 ★ 173repo stars claude-agent-sdk Guides Claude Code in building autonomous agents with hooks, subagents, and in-process MCP tools via Claude Agent SDK. CX Codex +1 ★ 173repo stars storytelling-with-data storytelling-with-data is a Claude Code skill for shaping data stories and reviewing visualisations. It kicks in when a chart, dashboard, infographic or report needs a clearer message. CX Codex +1 ★ 173repo stars extract-wisdom extract-wisdom turns a YouTube video, an article or a plain text file into key points, quotes and next steps. Handy when the source is long and you only need what matters. CX Codex +1 ★ 173repo stars

Project sponsors

Companies that keep this analytics open