pitch-deck-writer
Skill0 starsChanged 50 days agoMITView on GitHub
Builds an investor or lender pitch deck as a PowerPoint (.pptx) file from the business's feasibility study. It follows the proven narrative (problem, solution, why now, market, competition, business model, traction, financials, team, the ask), puts one idea per slide with the takeaway as the headline, fills every financial figure from Foundation's calculation engine through the bundled renderer, and runs visual QA on every slide. Use when someone asks for "a pitch deck", "investor presentation", "slides for my bank meeting", "turn my business plan into a deck", or a ".pptx". Works with a connected Foundation MCP server (engine-computed financial slides) or standalone (no financial result slides).
2 other projects use a version of this.
Adopted by
Most recently taken up by
- unrealandychan/AgentFoundrytheir own version, first seen Apr 22, 2026
- xcrrr/claude-skillstheir own version, first seen Mar 19, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- xcrrr/claude-skills★ 6same name
- unrealandychan/AgentFoundry★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- business-plan-writerSkill
- feasibility-analystSkill
- market-researchSkill
- pricing-optimizerSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 11 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Never type a computed figure.
- Distinctive: line 26
Look for the Foundation MCP tools (a client may add a prefix).
- Distinctive: line 32
Standalone: no engine figures. Build the story slides.
- Distinctive: line 36
Ask in one round for anything you still need: the audience (angel, VC, bank or grant body), the amount and instrument, the team bios, any …
- Distinctive: line 42
Write every slide's headline as its takeaway, as a full sentence.
- Very specific: line 68
Grant body: add impact, such as jobs (project.overview.employment) and local benefit.
- Distinctive: line 77
Every slide already carries a visual element: a big number, a chart, a grid, bars, or an accent rule.
- Distinctive: line 79
Use-of-funds percentages must add up to 100 (the renderer checks this).
- Very specific: line 82
Writing-style pass (required): Follow references/writing-style.md, which is based on Wikipedia's "Signs of AI writing" guide.
- Distinctive: line 99
Look at every slide PNG: Treat QA as a bug hunt, not a formality.
- Distinctive: line 102
If LibreOffice isn't available, say that visual QA wasn't possible, and still fix every renderer WARNING.
- Distinctive: line 113
Gemini: import into Google Slides and use Gemini there.
- Distinctive: line 138
No fabricated traction, logos, testimonials or team credentials.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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 Deck Writer
A seed investor spends about 3–4 minutes on a deck. Many aren't read to the end, and the first slide gets far more attention than any other. So the deck has to make its case in the headlines alone. YC's rules are legibility, simplicity and obviousness: one idea per slide, large type, and no decoration that doesn't carry meaning. Sequoia's structure is the default narrative, and it expects a bottom-up market size.
How numbers work
Same rule as the business-plan-writer skill:
- Computed figures appear in the spec only as
{{path|format}}tokens, filled from a facts file holding the engine's own output. - The projection chart (
"from": "results.projection"), the scenarios table and the competitor grid are generated from the facts directly. - Never type a computed figure. A path that doesn't exist stops the build.
Step 0: Mode and inputs
Look for the Foundation MCP tools (a client may add a prefix).
- Connected: confirm the project. Save
facts.jsonwithcurrency,project(get_project),results(compute_feasibility) andscenarios(compute_scenarios), pasted in verbatim. Run the market-research skill first if there's nocompetitorBenchmarkor market sizing yet. - Standalone: no engine figures. Build the story slides; for the financial slide, show the investment inputs and the pricing and say projections come from Foundation. Don't invent profit, IRR or payback figures.
Ask in one round for anything you still need: the audience (angel, VC, bank or grant body), the amount and instrument, the team bios, any traction (customers, pre-orders, letters of intent, revenue so far), and a one-line tagline.
Step 1: Storyline (write the headlines first)
Write every slide's headline as its takeaway, as a full sentence. For example, "No one on the route combines specialty coffee, food and a 7am opening", not "Competition". Read the headlines in sequence. If they don't tell the whole story alone, fix them before designing anything.
Default sequence (10–13 slides):
| # | Slide | Type |
| 1 | Company + one-line tagline | title |
| 2 | Problem: specific, quantified, from the customer's point of view | statement (+ points) |
| 3 | Solution / product | statement or bullets |
| 4 | Why now | statement |
| 5 | Market: bottom-up TAM/SAM/SOM with sources | market |
| 6 | Competition: the benchmark grid, your row highlighted | benchmark |
| 7 | Business model: pricing and unit economics (inputs, or saasMetrics tokens) | kpis or table |
| 8 | Traction / validation (skip it rather than fake it) | kpis or bullets |
| 9 | Financials: headline figures | kpis with engine tokens |
| 10 | Projection | chart from results.projection |
| 11 | Downside: Worst / Base / Best | scenarios |
| 12 | Team | team |
| 13 | The ask + use of funds + milestones | ask |
| 14 | Close / contact | closing |
- Bank or lender audience: move the downside and repayment capacity forward, and play down "why now" and exit.
- Grant body: add impact, such as jobs (
project.overview.employment) and local benefit.
Step 2: Write the spec
Format: references/spec-format.md.
- At most about 25 words of body text per slide, and 3–4 points maximum. Put detail in the
speaker notes (
notes). - Headlines: 1–2 lines. The renderer warns when text probably overflows its box.
- Every slide already carries a visual element: a big number, a chart, a grid, bars, or an accent rule. Don't add clip-art.
- Use-of-funds percentages must add up to 100 (the renderer checks this).
- Consistency: the same numbers as the business plan and the project. Same ask, same dates, same team.
- Writing-style pass (required). Follow
references/writing-style.md, which is based on Wikipedia's "Signs of AI writing" guide. No em dashes. Headlines must be plain claims, not slogans. If a humanizer skill built on that guide is installed (for exampleblader/humanizer), run it over the headlines, points and notes. It may change wording only, never a number, name, date, source or{{token}}. The renderer's style check must end with no style warnings.
Step 3: Render and QA
python scripts/build_pptx.py deck.json pitch-deck.pptx --facts facts.json --png qa/This needs python-pptx and matplotlib (preinstalled in Claude's sandbox), plus LibreOffice
and PyMuPDF for the PNG step when available.
- Exit code 2: a token path is wrong. Fix the path.
- Look at every slide PNG. Treat QA as a bug hunt, not a formality. Check for overflow, collisions, a headline that wraps to three lines, an unreadable chart, an empty-looking slide, or a number that contradicts another slide. Fix, re-render and look again.
- If LibreOffice isn't available, say that visual QA wasn't possible, and still fix every renderer WARNING.
Step 4: Optional second step: advanced restyling with another tool
The renderer produces a clean, correct deck. For the founder's own slide master or brand
template, native editable charts or heavier design, hand the finished .pptx to the
presentation tool their environment has, as a second step:
- Claude: Anthropic's
pptxskill. - ChatGPT / Codex: OpenAI's slides handling.
- Gemini: import into Google Slides and use Gemini there.
- Microsoft 365: Copilot and Designer in PowerPoint.
- Cursor and other agents: an open-source skill such as
ppt-master(MIT), which fills an existing template.
The table and links are in references/restyling.md. Tell that tool to change design
only. Afterwards run python scripts/check_numbers.py pitch-deck.pptx
pitch-deck-restyled.pptx, and look at the rendered slides again. If any number was lost,
changed or added, don't deliver the restyled file.
Step 5: Deliver
The .pptx, plus a one-paragraph note: the storyline in headline form, which slides rely on
[estimate] inputs, and what would strengthen the deck most (usually traction evidence).
References
references/spec-format.md: slide types and fields.references/writing-style.md: the no-AI-tells writing guide (Wikipedia's "Signs of AI writing", no em dashes).references/restyling.md: which tool to use for a second-step restyle, by environment, and how to verify the numbers survived.
Guardrails
- No fabricated traction, logos, testimonials or team credentials.
- The market numbers are the market-research skill's sourced figures, with sources on the slide.
- Competitor claims in the grid must be verifiable (public sources).
Sources
DocSend/Papermark deck-reading data (average review time about 3.7 minutes; first slide read
most; only about 58% read to the end), YC seed deck guidance (legibility, simplicity,
obviousness) and Sequoia's template: DocSend,
Papermark,
YC library,
Sequoia format.
Slide QA approach inspired by Anthropic's public pptx skill (referenced, not copied; that
skill is source-available, not open source).
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.

