pitchdeck-pitch-me
Skill0 starsChanged 50 days agoMITView on GitHub
Use BEFORE pitchdeck-grinding when you have a rough idea for a presentation and need to extract, question, and shape it first. Turns a vague idea into a confirmed brief with audience, goal, data, and assets mapped. Trigger on "pitchdeck-pitch-me", "pitch my idea", "I have an idea for a deck", "help me shape this presentation", or when no clear problem/goal/audience exists yet. Do NOT use when a completed pitch.md already exists — that starts pitchdeck-grinding.
In this repository9 other files it gives its agents.
- llms.txtllms.txt
- assets-generatorSkill
- audience-fitSkill
- brand-designSkill
- humanizerSkill
- pitchdeck-buildSkill
- pitchdeck-grindingSkill
- pitchdeck-reviewSkill
- pptxSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
When pitch.md exists and is confirmed → pitchdeck-grinding
- Very specific: line 23
When the user only needs a deck file rebuilt with no thinking → pitchdeck-build
- Very specific: line 27
If the user already hands you a complete brief (goal, audience, data, assets, constraints all stated), say so and offer: run the fast …
- Very specific: line 50
Check the workspace for existing material before greeting: prior decks (deck/*/), brand assets, data files, humanizer-context.md.
- Distinctive: line 61
Ask one question per message.
- Very specific: line 81
5W1H completeness check: after the walk-through: any W/H unaccounted?
- Distinctive: line 86
Run only if Phase C stays fuzzy after two attempts (user can't state the idea in one sentence) or the user asks for options.
- Very specific: line 118
Write deck/<slug>/pitch.md using the template from deckto init (run npx deckto init <slug> first if the workspace doesn't exist).
≈ “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
Pitch Me — idea extraction before any deck work
Extract the user's idea through a fixed questioning framework, converge it with
them, and write deck/<slug>/pitch.md. This skill and pitchdeck-grinding are
the pipeline's two core skills: this one gets the idea OUT of the user's head.
Core principle: Diverge first, converge second. No artifacts until the
framework is answered or explicitly marked unknown. Never invent answers.
When to Use
- "I have an idea for a deck / pitch / presentation"
- "Pitch my idea", "help me shape this", "I need slides for …" with no brief
- Idea exists but goal, audience, data, or assets are unclear
Do NOT use:
- When
pitch.mdexists and is confirmed →pitchdeck-grinding - When the user only needs a deck file rebuilt with no thinking →
pitchdeck-build
Off-ramp (advisory, non-blocking)
If the user already hands you a complete brief (goal, audience, data, assets,
constraints all stated), say so and offer: run the fast confirmation pass (Phase 1,
one question per gap only) or go straight to pitchdeck-grinding. Their call.
Hard Gates
GATE A: NO ARTIFACT WORK before Phases 1-3 complete.
pitch.md is written only at Phase 4. Silence is not an answer.
GATE B: EVERY framework slot (A-F) is filled OR marked `unknown`
in pitch.md. Never silently filled, never invented.
"Unknown" is a valid, honest answer — fabricated answers are not.
GATE C: USER MUST EXPLICITLY CONFIRM the converged synthesis.
"ok", "sure", or silence while scrolling is not confirmation.
GATE D: HANDOFF MUST PRESENT ALL 3 OPTIONS. Do not auto-invoke
pitchdeck-grinding on your own initiative.Phase 0 — Preflight (silent)
Check the workspace for existing material before greeting: prior decks
(deck/*/), brand assets, data files, humanizer-context.md. Greet with what
you found:
Found: <existing artifacts or "nothing yet">
What idea are we shaping?Phase 1 — Framework questioning (A–F)
Ask one question per message. Walk the framework in order; skip a slot only if the user already answered it. Use multiple choice when natural.
| Phase | Theme | Probe |
| A. Goal | decision & purpose | What's the goal? What should the audience do after? What decision are you asking for? |
| B. Audience | who & skepticism | Who's in the room? What do they already know? What's their likely objection? |
| C. Idea core | the content | The idea in one sentence? The problem? Why now? What evidence do you have? |
| D. Data readiness | evidence state | Is the data ready? Where from exactly? What's missing? May I research/estimate gaps (marked as such)? |
| E. Assets | existing material | Existing assets (brand kit, logos, prior decks, charts)? Or must they be sourced/generated? What theme/visual direction? |
| F. Constraints | format & stakes | Time budget? Slide count? Format (internal / pitch / keynote)? |
Questioning discipline (see references/question-framework.md for rationale +
sources):
- Indirect probing over accepting first answers: "You said the data exists —
where would someone actually find it?"
[EMPIRICAL: elicitation research] - SCQA framing for Phase A: push for Situation → Complication → Question →
Answer. If the user can't state a Complication, the idea isn't ready to grind —
say so plainly.
[BOOK/STANDARD: Minto] - 5W1H completeness check after the walk-through: any W/H unaccounted?
[CONVENTION] - Record every gap as
unknown— never fill it yourself.
Phase 2 — Diverge (conditional)
Run only if Phase C stays fuzzy after two attempts (user can't state the idea in one sentence) or the user asks for options.
Pick 3 methods from references/diverge-methods.md (default trio: Question
Storming, First Principles, Six Thinking Hats). Output each method's results
visibly:
### [Method] — [category]
[4-6 ideas generated against THIS idea]Then converge: present shortlist + discarded-with-reason.
Phase 3 — Converge + confirm
## Problem synthesis
Goal (SCQA): <S / C / Q / A, one line each>
Audience: <profile + primary objection>
Data: <ready | partial | missing> — <one line>
Assets: <existing | generate | research> — <one line>
Constraints: <time / slides / format>
Unknowns: <list — explicitly unknown, to be resolved or carried>
Does this match what you want? (I'll write it to pitch.md next)GATE C: explicit yes required. Revise on pushback.
Phase 4 — Write pitch.md + handoff
Write deck/<slug>/pitch.md using the template from deckto init (run
npx deckto init <slug> first if the workspace doesn't exist). Fill every
framework slot; put gaps under Unknowns — never fabricate.
Then GATE D — present exactly these three options:
pitch.md written to deck/<slug>/pitch.md
What next?
1. Invoke pitchdeck-grinding now (turns this into storyline + design spec)
2. Iterate on <phase> before proceeding
3. Save and stop hereOption 1 → invoke pitchdeck-grinding, passing the pitch.md path.
Option 2 → return to the named phase, revise, re-present the gate.
Option 3 → confirm path, stop cleanly.
Common Mistakes
| Mistake | Why it breaks the deck |
| Asking "what's your idea?" then immediately building | Produces the agent's idea, not the user's — this is the #1 baseline failure |
| Filling a framework slot by inference | GATE B violation; the fabricated goal steers every downstream slide |
| 15 unstructured questions in one message | Overwhelms the user, still misses phases — framework + one-per-message prevents both |
| Auto-invoking grinding after synthesis | User never approved; changes cost 4× more downstream |
| Skipping Phase E | Assets decide visual_source for every slide; guessing here means re-grinding later |
Reference Triggers
| Reference | When to Load |
references/question-framework.md | Phase 1: per-question rationale, SCQA probes, sources |
references/diverge-methods.md | Phase 2: method catalog + default trio |
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.

