investor-materials

Create and update pitch decks, one-pagers, investor memos, accelerator applications, financial models, and fundraising materials that stay internally consistent across every asset. Use when the user n

SKILL.md

GitHub

Investor Materials

Build investor-facing materials that are consistent, credible, and easy to defend.

When to use

  • creating or revising a pitch deck
  • writing an investor memo or one-pager
  • building a financial model, milestone plan, or use-of-funds table
  • answering accelerator or incubator application questions
  • aligning multiple fundraising docs around one source of truth

When NOT to use:

  • the ask is market sizing or competitor intelligence — that is market-research, which feeds this skill
  • the ask is the outreach email — that is investor-outreach
  • the canonical numbers do not exist or conflict. Drafting on top of conflicting numbers guarantees a partner meeting where two documents disagree; stop and resolve first

Inputs

The golden rule: all investor materials must agree with each other. Create or confirm one fact sheet before writing anything — conflicting numbers = stop and resolve, never draft over the conflict:

  1. Traction metrics — with the system they came from, not recollection. SkyyRose product, price, and SKU-count facts come from wordpress-theme/skyyrose-flagship/data/skyyrose-catalog.csv; live order/revenue facts come from the WooCommerce REST API (/wp-json/wc/v3, BasicAuth with keys in .env.wordpress) — the MCP wc_* tools have a broken auth username and return invalid_username.
  2. Pricing and revenue assumptions, each stated explicitly.
  3. Raise size and instrument.
  4. Use of funds — categories and percentages that sum to 100.
  5. Team bios and titles.
  6. Milestones and timelines.

Write the fact sheet to a real file (e.g. fact-sheet.md). Every downstream asset is checked against it mechanically, so it must exist as text, not as context.

Procedure

  1. Inventory the canonical facts into the fact sheet. Anything unknown is written as TBD (assumption owner: <name>), never as a confident number.
  2. Identify missing assumptions and mark which of them the story depends on.
  3. Choose the asset type and draft it with the logic visible, not buried:
    • Pitch deck — company + wedge · problem · solution · product/demo · market · business model · traction · team · competition/differentiation · ask · use of funds + milestones · appendix. For a web-native deck, pair with frontend-slides.
    • One-pager / memo — what the company does in one clean sentence · why now · traction and proof points early · a precise ask · claims easy to verify.
    • Financial model — explicit assumptions · bear/base/bull where useful · layer-by-layer revenue logic · milestone-linked spending · sensitivity analysis where the decision hinges on an assumption.
    • Accelerator application — answer the exact question asked · lead with traction, insight, team advantage · no puffery · metrics identical to the deck and model.
  4. Cross-check every number in the draft against the fact sheet — run the checks below, do not eyeball it.
  5. Re-run all checks after any fact-sheet edit. One changed metric invalidates every asset that quoted it.

Verification

Every check runs against files, so it can return "no". A check that errors on a missing file exits 2 — a dead gate, not a pass (bug-230); fix the path and re-run before reading the result.

  1. Use of funds sums to exactly 100 — the arithmetic a partner checks first:
python3 -c "import csv,sys; rows=list(csv.DictReader(open('use-of-funds.csv'))); s=sum(float(r['pct']) for r in rows); print('sum', s); sys.exit(0 if s==100 else 1)"

PASS: prints sum 100.0, exit 0. Any other total is a defect in the document, not a rounding footnote. [repro]

  1. Every number in every asset exists in the fact sheet — the golden rule, mechanised:
for n in $(grep -ohE '\$[0-9][0-9,.]*[KMkm]?|[0-9]+ (SKUs|customers|users)' deck.md memo.md | sort -u | tr ' ' '_'); do
  v=$(echo "$n" | tr '_' ' ')
  grep -qF "$v" fact-sheet.md || echo "MISSING from fact sheet: $v"
done

PASS: no MISSING lines. Each one is a number that exists in an investor-facing document and nowhere in the source of truth — resolve it in the fact sheet or delete it. [repro]

  1. Traction claims match the live system, not memory — for SkyyRose SKU counts:
python3 -c "import csv; print(sum(1 for _ in csv.DictReader(open('wordpress-theme/skyyrose-flagship/data/skyyrose-catalog.csv'))), 'catalog rows')"

PASS: the printed count equals the SKU count claimed in the deck. A mismatch means the deck is quoting a remembered number. [repo] — and note the scope: a catalog row is not proof the product is live on the storefront. A "live on skyyrose.co" claim needs its own [live] probe (curl -s "https://skyyrose.co/shop/?cb=$(date +%s)"), and without it the claim is downgraded, not asserted (bug-287).

Check 1 was proven able to fail before being trusted (rule 3) — see the worked example. A SKIP is not a PASS: whether the story is defensible in a partner meeting is a human judgment. State it as open and name who closes it; the mechanical greens only prove the numbers are internally consistent, not that the narrative holds.

Worked example

Real run, 2026-07-28, using files in the session scratchpad.

$ python3 -c "import csv,sys; rows=list(csv.DictReader(open('use-of-funds.csv'))); s=sum(float(r['pct']) for r in rows); print('sum', s); sys.exit(0 if s==100 else 1)" && echo "PASS: allocations sum to 100"
sum 100.0
PASS: allocations sum to 100

Allocation set: Inventory & production 40 · Paid acquisition 25 · Team 20 · Ops & tooling 10 · Buffer 5. The gate was proven red by adding 5 points to the total in the same run — it printed sum=105 and exited non-zero [repro].

Consistency check across memo.md (raising $500K on a SAFE, 33 SKUs live) against fact-sheet.md:

$ for n in $(grep -ohE '\$[0-9][0-9,.]*[KMkm]?|[0-9]+ SKUs' memo.md | sort -u | tr ' ' '_'); do v=$(echo "$n" | tr '_' ' '); grep -qF "$v" fact-sheet.md || echo "MISSING from fact sheet: $v"; done; echo "done exit=$?"
done exit=0

No MISSING lines — every figure in the memo traces to the fact sheet [repro]. Scope note on the 33 SKUs figure: the fact sheet records it as catalog rows [repo]. Any deck line claiming they are all purchasable on skyyrose.co needs its own [live] probe before it ships.

Failure modes

Failure What it looks like Rule
Documents disagree deck says 40 customers, memo says 35 The golden rule. Resolve in the fact sheet, then re-run check 2 on every asset
Use of funds does not sum allocations total 105% Check 1. It is the first thing a partner adds up
Metric quoted from memory SKU or revenue number with no system behind it bug-096 shape. Catalog CSV / WC REST are the SOT; plausible ≠ sourced
Repo → live scope jump "33 products live on the site" proven only by a CSV bug-287. [repo] does not cover a [live] claim. Probe or downgrade the wording
Broken tooling read as truth WC MCP wc_get_products returns invalid_username / empty and is treated as "no orders" bug-230 fail-open. The MCP auth path is known broken — use direct BasicAuth REST with .env.wordpress keys
Fuzzy market sizing TAM with no stated assumptions Every leap in logic gets an explicit assumption line; source the inputs via market-research
Inflated certainty fragile assumptions written as facts Show the assumption, show the sensitivity. Label inference as inference
Stale asset after a fact change one metric updated, three docs still quote the old one Re-run all checks after any fact-sheet edit
Mechanical greens read as a full pass "materials verified" with no defensibility read A SKIP is not a PASS — name the open judgment and its owner

Related skills

market-research (sizing, competitors, investor diligence) · skyyrose-market:investor-outreach (the emails; plugin-namespaced — verified on disk at ~/.claude/plugins/cache/skyyrose-suite/skyyrose-market/1.0.0/skills/, not a bare investor-outreach in .claude/skills/) · frontend-slides (web-native deck)