Skip to main content
skill CX Codex

seed-model — Claude Code skill: model and install

seed-model — a Claude Code skill that builds a 36-month financial model from business assumptions. It runs for questions about runway, break-even, fundraising, or before /seed-plan and /seed-deck.

MIT ★ 197 updated 2026-11-01

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add Jakeschincariol/seed-skill
/plugin install seed-skill@seed-skill
Open on GitHub

What seed-model does

The skill calculates customers, orders, GMV, revenue, costs, and cash month by month, then reports runway, the break-even month, peak burn, and unit economics. It also checks whether the planned use of funds matches the raise.

The model.py script reads seed/assumptions.json; init can create that file from the example. Run run --write to save results to seed/model.json, or csv to create seed/model.csv. Assumptions cover pricing, CAC, churn, payroll, operating costs, and one-off expenses.

Running the script requires python3. If the use of funds differs from the raise by even a cent, funds reports the gap and exits with code 1.

What it helps with

  • build a 36-month projection of revenue and cash
  • estimate how long the raised cash will last
  • identify break-even and the largest monthly loss
  • calculate CAC payback and LTV:CAC
  • check whether planned spending matches the raise

Example prompts for the agent

  • Build a 36-month financial model from seed/assumptions.json.
  • Calculate runway and the month the business breaks even.
  • Check the use of funds against the raise and export the model as CSV.

How to install seed-model

1Plugin marketplace

/plugin marketplace add Jakeschincariol/seed-skill
/plugin install seed-skill@seed-skill

This installs the whole seed-skill plugin with all its skills

2Skills CLI (Vercel)

npx skills add Jakeschincariol/seed-skill --skill seed-model -a claude-code

3Manually: put the files in place

~/.claude/skills/seed-model/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/seed-model/

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

Questions about seed-model

How do I install seed-model in Claude Code?

Run /plugin marketplace add Jakeschincariol/seed-skill, then /plugin install seed-skill@seed-skill.

What does seed-model do?

It builds a 36-month financial model and calculates runway, break-even, and unit economics. It can also check the use of funds and save results as JSON or CSV.

What does seed-model need to run?

The script requires python3. Its calculations use the assumptions in seed/assumptions.json.

Source text

SKILL.md · MIT
---
name: seed-model
description: >-
  Financial Modeler. Builds the 36-month financial model behind your raise:
  customers, orders, GMV, revenue (GMV x your take rate), contribution,
  payroll, marketing (new customers x CAC), opex, net and cash, month by
  month, then the numbers investors ask for: runway, the breakeven month,
  peak burn, revenue for years 1 to 3, and the unit economics (contribution
  per order, CAC payback, LTV:CAC). Checks that the use of funds adds up to
  the raise and exports a CSV. Reads seed/assumptions.json, writes
  seed/model.json for the business plan, the deck and the pitch coach. Use
  when the user says "build my financial model", "how long will the money
  last", "what's my runway", "when do we break even", "3-year projections",
  "use of funds", "how much should I raise", "unit economics", "burn rate",
  or before /seed-plan and /seed-deck.
argument-hint: "[run | funds | csv] [--write]"
---

# seed-model

The Financial Modeler turns a handful of assumptions into the 36 months an
investor will ask about. You say what a walk (or an order, or a seat) costs,
how much of it you keep, how many customers you win and lose, who you hire
and when, and how much you are raising. It does the arithmetic, every month,
to the cent, and tells you when the money runs out, when the business pays
for itself and what each customer is worth. Every number traces back to an
assumption you can change and defend. It is a plan, not a forecast: the
model is only as honest as its inputs.

`${CLAUDE_SKILL_DIR}` below is this skill's folder (where this SKILL.md lives).

```bash
python3 "${CLAUDE_SKILL_DIR}/model.py" init                # seed/assumptions.json from the example
python3 "${CLAUDE_SKILL_DIR}/model.py" run                 # the summary, nothing written
python3 "${CLAUDE_SKILL_DIR}/model.py" run --write         # also writes seed/model.json
python3 "${CLAUDE_SKILL_DIR}/model.py" run --months        # every month in a table
python3 "${CLAUDE_SKILL_DIR}/model.py" funds               # the use of funds against the raise
python3 "${CLAUDE_SKILL_DIR}/model.py" csv                 # seed/model.csv for a spreadsheet
```

Other flags: `--dir PATH` (default `./seed`), `--assumptions PATH`, `--json`,
`csv --out -` to print the CSV.

## What it reads and writes

| file | what |
| --- | --- |
| `assumptions.example.json` (this folder) | the Leashly example `init` copies: every field filled in |
| `seed/assumptions.json` | your assumptions: price, take rate, cost per order, customers, CAC, churn, payroll, opex, one-off costs, the raise, the use of funds |
| `seed/model.json` | with `run --write`: the assumptions, 36 monthly rows, the summary and the unit economics. seed-plan, seed-deck and seed-coach read it |
| `seed/model.csv` | with `csv`: the 36 rows, one column per line item |

## The assumptions

| field | what it means |
| --- | --- |
| `price` | what a customer pays for one unit (a $25 walk) |
| `take_rate_pct` | the share of that price you keep (15 means 15%) |
| `cost_per_order` | what one unit costs you: `pct_of_price` and/or `fixed`, like card fees (2.9% + $0.30) |
| `customers` | `starting`, `new_month_1`, `new_growth_pct` (one monthly rate a year), `churn_monthly_pct`, `orders_per_customer_month` |
| `cac` | what it costs to win one new customer; marketing = new customers x CAC |
| `payroll`, `payroll_overhead_pct` | each role's monthly pay and start month; overhead covers payroll taxes and benefits |
| `opex`, `one_off` | monthly costs (software, insurance) and one-time costs in a given month (a city launch) |
| `raise`, `raise_month`, `cash_at_start` | the round, the month it lands, cash before it |
| `use_of_funds` | `{what: amount}`, adding up to the raise |

## The method

Each month: customers = last month's minus churn plus the new ones. Orders =
customers x orders a month. GMV = orders x price. Revenue = GMV x take rate.
Gross profit (the contribution) = revenue minus the per-order costs. Opex =
payroll with overhead + marketing + monthly opex + any one-off cost. Net =
gross profit minus opex. Cash = last month's cash + net (+ the raise when it
lands). Customers and orders round half up to whole numbers; money is exact
to the cent.

- **Breakeven month**: the first month with net at or above $0 that every
  later month keeps. A single good month does not count.
- **Peak burn**: the biggest monthly loss, and the month it happens.
- **Runway**: the months the cash covers before it first goes below $0. If
  it never does in 36 months, the summary says the cash lasts all 36.
- **Unit economics**: revenue per order = price x take rate. Contribution
  per order = that minus the costs of one order. CAC payback = CAC /
  (contribution per order x orders a month). LTV = monthly contribution per
  customer / monthly churn. LTV:CAC = LTV / CAC.
- **Use of funds**: must add up to the raise to the cent, or `funds` exits 1
  and says by how much it is short or over.

## Steps

1. If `seed/assumptions.json` does not exist, run `model.py init` and walk
   the founder through every field in plain words. Replace the Leashly
   numbers; never keep an example number the founder did not confirm.
2. Ask where each number comes from (a pilot, a quote, a competitor's public
   price). Note the source next to it in the chat. Guesses are fine if they
   are called guesses.
3. Run `model.py run`. Read back the yearly table, the unit economics,
   breakeven, peak burn and runway.
4. If the cash runs out before breakeven, say so plainly and show the two
   levers: raise more, or cut the plan (slower hiring, lower CAC). Never
   tweak an assumption on your own to make the picture prettier.
5. Run `model.py funds` until the use of funds adds up to the raise.
6. Run `model.py run --write`, then `model.py csv` if they want a
   spreadsheet.
7. Hand off to `/seed-plan` (it pulls these numbers into the business plan),
   then `/seed-deck`. `/seed-coach` will ask about every number here.

## Example: the reel

Leashly (fictional): $25 flat walks, a 15% take rate, card processing 2.9% +
$0.30 and a $0.50 backup-walker standby fee per walk, 210 pilot owners, 150
new owners in month 1 growing 10% a month in year 1, 7% monthly churn, 8
walks per owner a month, a $60 CAC, a $1.5M seed landing in month 1 (Jan
2027). `model.py run`:

```
Leashly: 36-month model, Jan 2027 to Dec 2029 (USD)
                             Year 1         Year 2         Year 3
  Walks                     124,640        423,104        798,272
  GMV                    $3,116,000    $10,577,600    $19,956,800
  Revenue                  $467,400     $1,586,640     $2,993,520
  Gross profit             $277,324       $941,406     $1,776,155
  Opex                     $957,800     $1,467,120     $1,679,880
  Net                     -$680,476      -$525,714        $96,275
  Cash at year end         $819,524       $293,810       $390,086
  Customers (end)             2,490          6,200          9,999

  Per walk           price $25.00, you keep $3.75 (15%), costs $1.53, contribution $2.23
  Per customer       LTV $254.29 on a $60.00 CAC: LTV:CAC 4.24x, CAC paid back in 3.37 months
  Breakeven          month 29 (May 2029), and every month after it
  Peak burn          $82,801 in month 5 (May 2027)
  Runway             the cash lasts all 36 months (low point $252,553 in month 28)
  Use of funds       adds up to the $1,500,000 raise
```

Peak burn lands in month 5 because the second city's $25,000 launch cost
hits that month. The use of funds: product and engineering $600,000 (40%),
city launches $450,000 (30%), marketing $300,000 (20%), operations and
insurance $150,000 (10%).

## Rules

- Every number comes from `seed/assumptions.json`. Nothing is looked up or
  invented by the tool.
- Projections from your own assumptions, not a forecast and not a promise of
  funding. Not investment advice.
- Show the bad months as clearly as the good ones: burn, the low point, a
  runway that ends.
- Never change an assumption to hit a target without saying so and getting
  the founder's yes.

© Jake Schincariol · MIT · Jakeschincariol/seed-skill

Similar skills

All skills for Finance and business →

From the same set12

The whole set →
seed-updates The seed-updates Claude Code skill drafts monthly investor updates with metric changes and a runway calculation. It can also prepare a shorter version for prospects who previously said “not now.” CX Codex +1 ★ 197repo stars seed-deck The seed-deck Claude Code skill turns a business plan into a 10-slide pitch deck. It creates an HTML presentation and checks its content and figures. CX Codex +1 ★ 197repo stars seed-terms The seed-terms Claude Code skill calculates ownership and dilution from SAFE and priced-round terms. It runs when asked to assess a deal or flag clauses for legal review. CX Codex +1 ★ 197repo stars seed-plan The seed-plan Claude Code skill builds a nine-section business plan from a startup idea. It runs when asked to draft or check a plan, work out unit economics, or prepare for /seed-deck. CX Codex +1 ★ 197repo stars seed-scout The seed-scout Claude Code skill examines competitors’ App Store reviews and groups recurring complaints. It runs for competitor research based on an idea or before drafting a plan and pitch deck. CX Codex +1 ★ 197repo stars seed-investors The seed-investors Claude Code skill finds investors for a startup and scores their fit. It runs on requests to find or rank investors, after /seed-plan and before /seed-closer. CX Codex +1 ★ 197repo stars seed-closer The seed-closer Claude Code skill drafts personalised investor emails and tracks replies through to a booked call. It responds to requests to send outreach, sort replies, or arrange a meeting. CX Codex +1 ★ 197repo stars seed-skill This Claude Code plugin brings together 11 skills for preparing a startup to raise a seed round. They cover competitor research, investor outreach, financial planning and pitch preparation. ★ 197repo stars

Project sponsors

Companies that keep this analytics open