pm-storytelling
Skill1 starsChanged 42 days agoMITView on GitHub
Turn raw PM source material — discovery transcripts, evidence dossiers, briefs, AI-generated drafts, scattered notes, assignments — into audience-ready narrative artefacts. Use whenever the output needs a spine (tension → insight → change → takeaway) rather than a flat summary. Covers PM-native formats (one-pager opener, PRD problem statement, exec / decision memo, release notes, discovery synthesis, QBR / slide-deck storyline with an optional .pptx render, customer case study, pitch deck) and generic prose (essay, article, video script, lesson, social sequence). Trigger on "rascunha a narrativa", "monta a história", "esse texto está sem alma", "give this a spine", "make this memorable" — the full trigger list lives in the skill body. Pushy by design — when source material has a real tension/insight/change to surface, prefer this over plain summarisation. If the request is to tighten prose without changing its shape, use humanizer instead.
In this repository65 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- anti-slopSkill
- data-science-analystSkill
- humanize-deliverablesSkill
- humanizerSkill
- inference-disciplineSkill
- pm-archetype-aiSkill
- pm-archetype-enterpriseSkill
- pm-archetype-growthSkill
- pm-archetype-platformSkill
- pm-phase-defineSkill
- pm-phase-deliverSkill
- pm-phase-developSkill
- pm-phase-discoverSkill
- pm-prioritization-regua-comumSkill
- pm-product-senseSkill
- pm-storytellingSkill
- pm-transversal-analysisSkill
- pm-transversal-commsSkill
- pm-transversal-docsSkill
- pm-transversal-stakeholderSkill
- repo-doctorSkill
- anti-slopSkill
- data-science-analystSkill
- humanize-deliverablesSkill
- humanizerSkill
- inference-disciplineSkill
- pm-archetype-aiSkill
- pm-archetype-enterpriseSkill
- pm-archetype-growthSkill
- pm-archetype-platformSkill
- pm-phase-defineSkill
- pm-phase-deliverSkill
- pm-phase-developSkill
- pm-phase-discoverSkill
- pm-prioritization-regua-comumSkill
- pm-product-senseSkill
- pm-transversal-analysisSkill
- pm-transversal-commsSkill
- pm-transversal-docsSkill
- pm-transversal-stakeholderSkill
- repo-doctorSkill
- umbertoSkill
- anti-slopSkill
- data-science-analystSkill
- humanize-deliverablesSkill
- humanizerSkill
- inference-disciplineSkill
- pm-archetype-aiSkill
- pm-archetype-enterpriseSkill
- pm-archetype-growthSkill
- pm-archetype-platformSkill
- pm-phase-defineSkill
- pm-phase-deliverSkill
- pm-phase-developSkill
- pm-phase-discoverSkill
- pm-prioritization-regua-comumSkill
- pm-product-senseSkill
- pm-storytellingSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
19 rules. About 0 are common, 16 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈16
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Use this skill when the user gives or references:
- Very specific: line 55
If the request is "tighten this prose" without "give it a spine" — use humanizer directly, not this skill.
- Distinctive: line 67
Default chain when the artefact ships outside the immediate team:
- Distinctive: line 78
No empty drama: Every scene, hook, metaphor, and example must support the goal.
- Very specific: line 79
Evidence anchored: Do not invent facts, data, citations, claims, company results, customer quotes, or personal experiences.
- Distinctive: line 81
One strong spine: Every output needs a single sentence that survives the test below.
- Distinctive: line 87
Every generated output must have one sentence that fits this template:
- Distinctive: line 91
If the spine is weak, refine it before drafting.
- Distinctive: line 109
If the user already provided enough information, proceed without asking.
- Distinctive: line 125
Before writing, extract three layers (briefly, in scratchpad form is fine — the user does not need to see them unless asked):
- Distinctive: line 155
When source is Thin / Conflicted / Unsupported, proceed with placeholders or push back.
- Distinctive: line 181
Before final writing, produce a compact plan when useful (skip only if the user said "just the final text"):
- Very specific: line 202
Use the most relevant template from templates/OUTPUT_TEMPLATES.md.
- Distinctive: line 207
Use specific nouns, verbs, examples, and constraints from the source
- Distinctive: line 210
Avoid moralising, clichés, and vague transformation language
- Distinctive: line 212
Use headings only when they aid readability or match requested format
- Distinctive: line 213
Do not invent metrics, testimonials, citations, names, events, dates, or outcomes
- Distinctive: line 229
Before final response, check the following.
- Distinctive: line 248
Use this default unless the user requests otherwise:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
pm-storytelling: turn source material into narrative without losing the assignment
Forked from the standalone assignment-storytelling skill and adapted for the PM skills repo. The job is the same: take an assignment, brief, evidence pile, or rough draft and produce a clear, compelling narrative that preserves the original intent. The change is the lane: this skill generates narrative; voice polish and outbound delivery are someone else's job in this repo (see next section).
Trigger phrases
The frontmatter description carries a subset; this is the full list. Invoke on: "rascunha a narrativa", "monta a história", "transforma essas notas em narrativa", "story for the exec memo", "PRD opener", "one-pager narrative", "release notes story", "discovery synthesis story", "case study from this customer", "pitch storyline", "slide narrative", "monta o deck", "QBR deck", "gera o pptx", "make this less generic", "give this a spine", "esse texto está sem alma", "make this memorable".
Progressive loading
Load this SKILL.md first. For large or specialized tasks, use references/progressive-loading.md to choose the narrowest supporting reference before reading more.
Why this exists
A lot of PM output reads like an assembled report — facts stacked on facts, no spine, no point of view. Storytelling adds a spine. But "more story" is not the goal — clarity is. Every scene, hook, metaphor, and example must serve the assignment, not decorate weak content. When source material is thin, the right move is to mark gaps explicitly ([NEEDS SOURCE]), not invent.
When to use / when not
Use this skill when the user gives or references:
- a discovery synthesis, evidence dossier, or interview transcript pile that needs a narrative spine
- a one-pager / PRD / decision memo that needs a problem-statement opener with tension
- AI-generated content that reads flat, generic, or disconnected
- raw notes, brainstorms, transcripts, product notes, or lesson material
- a request to make content more narrative, persuasive, educational, memorable, or presentation-ready
- an assignment, exercise, brief, prompt, or rubric (the legacy use case)
Skip this skill for:
- pure copy-editing where the user wants the same shape, just cleaner
- structured ticket fields (Jira labels, components, fixVersion)
- raw machine output — JSON, logs, CSVs, terminal commands
- code, config, schemas
- internal scratchpads or
.ai/memory/updates that nobody outside the team will read
If the request is "tighten this prose" without "give it a spine" — use humanizer directly, not this skill.
Working with humanizer and humanize-deliverables
This skill always pairs with humanizer for voice polish, and with humanize-deliverables when the artefact ships outbound (Confluence, Slack, customer, leadership). Storytelling and humanizing are distinct jobs in this repo. Don't merge them.
| Skill | Role | When |
pm-storytelling (this) | Generate narrative — spine, beats, format | Source needs structure or arc |
humanizer | Polish voice — strip AI-writing patterns | Voice quality matters (long-form, exec-facing, customer-facing) |
humanize-deliverables | Gate outbound prose — hard-enforced sentinel | Before Confluence / Slack / customer / exec / release notes |
Default chain when the artefact ships outside the immediate team:
- Draft here — produce the story (Phases 0–6 below).
- Polish with
humanizer— read../humanizer/SKILL.md, apply its pattern catalogue, and run its mandatory final pass ("what still makes this AI-generated?"). Don't reimplement that catalogue inside this skill — the inline anti-generic filter (Phase 5) is a tiny first sweep, not a substitute. - Run
humanize-deliverablesbefore delivery — read../humanize-deliverables/SKILL.md. The gate is hard-enforced forcreateConfluencePage,updateConfluencePage,slack_send_message,slack_send_message_draft, Slack canvases, and Jira comments. Forgetting it blocks the call.
Skip the humanizer/gate pass for: internal scratchpads, .ai/memory/ updates, tickets you're triaging, raw analytical output, or anything you'll rewrite anyway. When unsure, run them — the cost of one extra pass is seconds; the cost of AI-tinted prose hitting a customer or a director is reputational drag.
Core principles
- Assignment first. Preserve the task objective, constraints, audience, and required deliverable before adding narrative polish.
- No empty drama. Every scene, hook, metaphor, and example must support the goal. PM storytelling is dragged forward by evidence, not by emotional arc.
- Evidence anchored. Do not invent facts, data, citations, claims, company results, customer quotes, or personal experiences. Mark gaps with
[NEEDS SOURCE: …]. - Audience fit. Adapt tone, structure, depth, and examples to the actual reader (evaluator, customer, eng director, GPM, learner). When in doubt, default to a lean house style: direct, evidence-first, occasionally bilingual PT/EN.
- One strong spine. Every output needs a single sentence that survives the test below.
- Reusable formats. The same source can become an essay, an article, a video script, a slide storyline, a case study, a one-pager, a memo, or a social sequence. Pick one primary format; flag others as adjacent assets if useful.
- Iterate visibly when source is messy. First extract ingredients, then propose the route, then write. Don't pretend a thin pile of notes is ready to ship.
The narrative-spine test
Every generated output must have one sentence that fits this template:
This is a story about [protagonist] who faces [friction], discovers [turning point], and ends with [takeaway].
If the spine is weak, refine it before drafting. PM-flavoured example:
This is a story about active customers who plateau after 90 days, the team that mistook adoption for stickiness, the discovery that drove churn was account-management drift, and the bet that closes that gap.
Phase 0 — detect mode
| Mode | When the user gives you… | Goal |
| A — Assignment-to-Story | An assignment, prompt, rubric, brief, or task | Satisfy the task while making the answer narrative |
| B — Generated-to-Story | AI-generated or drafted content | Preserve substance, strip generic phrasing, add spine |
| C — Raw-Notes-to-Story | Notes, bullets, transcripts, fragments | Extract spine, organise material |
| D — Story Adaptation | An existing story that needs a different format / channel / audience / length | Preserve story core, change delivery |
| E — Multi-Asset System | Source + a request for several outputs (article + slide storyline, script + social posts, case study + exec summary, lesson + activity) | Consistent story world across assets |
Phase 1 — intake
If the user already provided enough information, proceed without asking. Ask at most three objective questions, only when critical info is missing:
- Target format — essay, article, script, slide storyline, one-pager, PRD opener, exec memo, release notes, case study, lesson, social sequence
- Audience and tone — who reads it, what register
- Constraints — word count, language (PT / EN / both), rubric, citations, deadline, brand voice, forbidden claims, required sections
Defaults when nothing is specified:
- Format: narrative article or structured story draft
- Language: match the user's language; if the user mixes PT/EN in the source, mirror that
- Tone: clear, professional, direct
- Length: concise but complete
- Claims: only what's in the provided material; mark unknowns as
[NEEDS SOURCE]
Phase 2 — source deconstruction
Before writing, extract three layers (briefly, in scratchpad form is fine — the user does not need to see them unless asked):
Assignment map
- Objective — what the task actually asks for
- Deliverable — required format and scope
- Audience / evaluator — who judges success
- Required criteria — rubric, learning outcomes, business goals, constraints
- Non-negotiables — topics, sources, terms, frameworks, citations, length
- Risk — where storytelling could accidentally violate the assignment
Story ingredients
- Protagonist — person, team, company, customer, learner, idea, problem, community
- Desire — what the protagonist wants
- Friction — obstacle, misconception, constraint, tension, unanswered question
- Stakes — why it matters now
- Turning point — insight, discovery, decision, experiment, conflict, shift
- Resolution — result, lesson, recommendation, next step
- Takeaway — what the audience should remember or do
Content quality check
| State | What it means | What to do |
| Ready | Enough detail to write a final story | Draft the final output |
| Thin | Usable but lacks examples, proof, stakes | Draft with placeholders; flag what's missing |
| Conflicted | Unclear claims, inconsistent structure, mixed objectives | Surface the conflict to the user; ask to resolve before drafting |
| Unsupported | Requires facts, citations, or data not provided | Refuse to invent; return a gap list |
When source is Thin / Conflicted / Unsupported, proceed with placeholders or push back. Inventing facts to round out the arc is the exact failure mode this skill exists to prevent.
Phase 3 — choose narrative architecture
Pick one primary framework. See frameworks/STORY_FRAMEWORKS.md for the full catalogue (13 frameworks).
Default choices:
| Situation | Framework |
| Discovery synthesis | Question → Method → Finding → Pattern → Implication → Open |
| One-pager opener | Why now → What we found → What we'll do → Sizing → Risk |
| PRD problem statement | Customer + JTBD → Friction observed → What we tried → Why now → Bet |
| Exec / decision memo (DACI) | Decision → Why now → Context → Options weighed → Choice + reasoning → Acknowledged risk |
| Release notes | What changed for the user → Why it matters → How to use → What's next |
| Persuasive recommendation | Problem → Stakes → Evidence → Recommendation |
| Customer case study | Context → Challenge → Approach → Evidence → Learning → Next |
| Slide / QBR storyline | Promise → Problem → Stakes → Insight → Solution → Proof → Implications → Action; per-slide contract and deck rules in references/deck-storyline.md |
| Pitch | Why now → Pain → Shift → Solution → Proof → Ask |
| Academic / business analysis | SCQA + Insight Arc |
| Personal reflection | Moment → Conflict → Realisation → Change |
| Educational | Curiosity Gap → Explanation → Example → Application |
| Short-form video / social | Hook → Tension → Reveal → Payoff → CTA |
Phase 4 — output planning
Before final writing, produce a compact plan when useful (skip only if the user said "just the final text"):
## Story strategy
- Objective:
- Audience:
- Format:
- Narrative spine:
- Chosen framework:
- Key beats:
1.
2.
3.
- Evidence available:
- Gaps / placeholders:This makes assumptions visible and lets the user redirect cheaply before bigger writing happens.
Phase 5 — generate the story
Use the most relevant template from templates/OUTPUT_TEMPLATES.md.
Universal output rules
- Open with a concrete hook, not a generic introduction
- Use specific nouns, verbs, examples, and constraints from the source
- One idea per paragraph
- Show tension early
- Avoid moralising, clichés, and vague transformation language
- Preserve assignment requirements even when improving style
- Use headings only when they aid readability or match requested format
- Do not invent metrics, testimonials, citations, names, events, dates, or outcomes
- Mark missing facts as
[NEEDS SOURCE: …] - End with a clear takeaway, recommendation, reflection, or call to action
Inline anti-generic check (light)
While drafting, watch for the most flagrant generic openings: "in today's fast-paced world", "this essay will discuss", "throughout history", "in the rapidly evolving landscape of", "at its core", "it is crucial to note that", "a game-changer", "unlock the potential". If one slips in, replace with concrete context, tension, or a direct claim.
This is a first sweep, not a complete check. The full pattern catalogue lives in humanizer/SKILL.md and its references/; run that skill on the draft when voice quality matters, and always before outbound delivery.
Phase 6 — format-specific rules
Pick the deliverable format and load two files: templates/OUTPUT_TEMPLATES.md for the fillable structure, and references/format-editorial-rules.md for the per-format editorial rules (what to say once the template is in front of you — covers essay, one-pager, exec memo, release notes, article, video script, slide storyline, case study, lesson, social sequence). When the deliverable is a deck (QBR, exec review, stage-7 kickoff, pitch), also load references/deck-storyline.md for the per-slide contract, the slide budget, and the optional render handoff.
Phase 7 — quality review
Before final response, check the following. If any check fails, revise — don't ship.
- Assignment alignment — does it answer the original task and respect non-negotiables?
- Narrative spine — is the protagonist / friction / turning point / takeaway sentence clear?
- Specificity — are there concrete details from the source, not generic statements?
- Evidence integrity — are unsupported facts marked, not invented?
- Audience fit — does tone match the intended reader?
- Format fit — does it follow requested structure and length?
- Memorability — at least one strong image, contrast, example, or sentence?
- Voice gate (outbound only) — has the draft been through
humanizerand, if it ships externally, marked viahumanize-deliverables?
The voice gate item is binary, not a score. If the artefact is outbound prose and hasn't been humanised, the work is not done.
See rubrics/QUALITY_RUBRIC.md for the 1–5 scoring scorecard with minimum bars.
Default response shape
Use this default unless the user requests otherwise:
## Story strategy
- Objective:
- Audience:
- Format:
- Narrative spine:
- Gaps / assumptions:
## Final story
[Generated output]
## Why this works
[2–4 bullets explaining alignment with assignment, narrative structure, and audience]For quick requests, drop "Why this works" and return only the final story. For outbound artefacts, append a one-line note: "Run humanizer on this draft before delivery; if it ships to Confluence / Slack / customer / exec, run humanize-deliverables first."
Support files
Load these only when needed; they're referenced from this SKILL.md:
frameworks/STORY_FRAMEWORKS.md— 13 narrative frameworks + selection heuristic (PM-specific frameworks added: Discovery Synthesis, PRD Opener, Decision Memo, Release Notes)templates/OUTPUT_TEMPLATES.md— fillable templates per format (PM-specific templates added)rubrics/QUALITY_RUBRIC.md— 1–5 scorecard across 7 dimensions plus a voice-gate pre-requisite for outboundexamples/sample_one_pager_opener.md— worked example showing assignment → strategy → final story → why-this-worksreferences/deck-storyline.md— per-slide assertion-evidence contract, QBR slide budget, and the optional harness-dependent.pptxrender handoff
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

