Business Plan skill for Codex
A plan short enough to update monthly. Codex starts every business plan as a one-page plan, names the riskiest assumptions with a cheap test and pass threshold for each, and only then expands into the format the reader needs, whether that is a Lean Canvas, an internal case or a bank-ready plan with cash flow and debt coverage. Every number comes from a linked model, and claims without evidence are marked as placeholders rather than invented.
curl -fsSL codexskills.dev/install.sh \ | sh -s -- business-planInstalls to ~/.agents/skills/business-plan/ for every project on this machine.
curl -fsSL codexskills.dev/cli.mjs -o codex-skills.mjsnode codex-skills.mjs install business-planThe CLI also searches, updates and removes skills. Run it with no arguments to pick from a list. Needs Node 20.12+.
curl -fsSL codexskills.dev/install.sh \ | sh -s -- --project business-planInstalls to .agents/skills/business-plan/ in the current repository. Commit it and everyone who runs Codex there gets it.
curl -fsSLO codexskills.dev/downloads/business-plan.zipunzip -o business-plan.zip -d ~/.agents/skills/The archive holds the skill folder with its generated agents/openai.yaml. Unzip into .agents/skills/ to scope it to a repo.
& ([scriptblock]::Create((irm codexskills.dev/install.ps1))) business-planPowerShell 5.1 or 7. Installs to %USERPROFILE%\.agents\skills\. Add -Project for the current repo.
104-line POSIX script. No sudo, no telemetry. Read install.sh
Use it in Codex
Start a prompt with $business-plan, or pick it from /skills.
You can also just describe the task: Codex reads the description below and loads the playbook when your request matches.
What Codex reads before loading it
Writes lean business plans that are short, testable and tied to numbers - a one-page plan or Lean Canvas, a 10-20 page plan for banks, partners or grant bodies, explicit riskiest assumptions with tests, milestones, and a financial summary linked to a model. Use when asked to write a business plan, a lean canvas, a plan for an SBA or bank loan, a plan for a new product line or side business, or to turn a rough idea into a structured plan with assumptions and milestones.
Use it when
- Turning a business idea into a structured plan with milestones
- A lender, partner or programme asks for a written business plan
- You want a Lean Canvas and a list of assumptions to test
- An existing plan needs its numbers checked against a forecast
Use something else when
- The audience is venture investors and you need slides: use pitch-deck
- You need the detailed monthly forecast itself: use financial-modeling
- You only need TAM, SAM and SOM sizing: use market-research
How to check the result
Every revenue, cost and headcount figure in the plan matches the linked model, each riskiest assumption has a dated test and pass threshold, and no claim lacks a source or placeholder.
The mistake it prevents
Writing thirty polished pages before testing whether a single customer will pay the stated price.
A useful plan is short enough to be revised monthly, names the assumptions that would kill the business if wrong, and connects every revenue or cost claim to a number in a model. Length is set by the reader: a founder needs one page, a bank needs evidence of repayment, a grant body needs its own headings.
Plans for loans, visas, grants or investors may carry legal or financial consequences. Say that an accountant should check the financials and, for visa or loan applications, that the specific programme’s requirements and an adviser take precedence over this template.
Choose the format
| Reader | Format | Length | What they look for |
|---|---|---|---|
| Founder / internal team | Lean Canvas or one-page plan | 1 page | Focus, assumptions to test, next milestones |
| Investors | Pitch deck plus short memo | 10-14 slides | Growth and scale (see pitch-deck) |
| Bank or SBA lender | Traditional plan | 15-25 pages | Cash flow coverage, collateral, owner experience |
| Grant or accelerator | Their template exactly | As specified | Fit with the programme’s criteria |
| New product line in a company | Internal business case | 2-6 pages | Investment, return, risks, decision needed (see decision-memos) |
Ask which reader before writing. If there is no external reader, default to the one-page plan; it is the version people actually update.
Workflow
- Interview the user, or read their notes, for: customer, problem, offer, price, channels, costs, team, money available, and what they believe that others do not.
- Write the one-page plan first, whatever the final format. Longer plans expand it; they never introduce claims the one-pager lacks.
- List the riskiest assumptions and a cheap test for each (below).
- Build or link a simple financial model (see financial-modeling) and pull the summary numbers from it, never type them twice.
- Expand into the target format, keeping every section tied to evidence.
- Run the consistency and evidence checks, then deliver with a list of open questions.
One-page plan
# <Business> - plan as of <date>
**Customer:** <specific segment, size of segment, where they are found>
**Problem:** <what it costs them today, in time or money>
**Offer:** <product or service, how it is delivered>
**Price and unit economics:** <price, cost to deliver, gross margin per unit>
**Channels:** <how the first 100 customers will be reached; cost per customer if known>
**Competition / alternatives:** <what customers do today and why they would switch>
**Unfair advantage:** <something hard to copy, or "none yet">
**Team:** <who does what; gaps>
**Money:** <cash available, monthly costs, months until break-even or next funding>
**Milestones (next 6-12 months):** <3-5 dated, measurable milestones>
**Riskiest assumptions:** <top 3, each with a test and a deadline>Lean Canvas (Ash Maurya’s adaptation of the Business Model Canvas) has nine boxes: problem, customer segments, unique value proposition, solution, channels, revenue streams, cost structure, key metrics, unfair advantage. Fill problem and customer segments first; solution last.
Riskiest assumptions
Write each as a falsifiable statement with a test that costs days, not months:
| Assumption | Why it matters | Test | Pass threshold | Deadline |
|---|---|---|---|---|
| Clinic managers will pay $150/month | Whole revenue model | Pre-sell to 20 contacted clinics | 4 paid deposits | 2026-12-15 |
| We can reach them for under $300 CAC | Payback under 6 months | $1,500 LinkedIn test campaign | 5+ qualified demos | 2026-11-30 |
| Onboarding takes under one hour | Support cost per customer | Concierge onboarding of 5 pilots | median under 60 min | 2027-01-10 |
Order by “most likely to be wrong x most damaging if wrong”. Test the top one first.
Traditional plan sections (bank, partner, programme)
- Executive summary - one page, written last, states the amount requested and how it will be repaid or used.
- Company description - legal structure, location, ownership, history.
- Market analysis - target segment size (bottom-up), local competitors, pricing evidence. Cite sources with dates; see market-research for sizing.
- Products and services - what is sold, pricing, suppliers, IP if any.
- Marketing and sales - channels, sales process, conversion assumptions.
- Operations - location, equipment, suppliers, staffing, key processes.
- Management and organisation - owners’ relevant experience, advisers, hires.
- Financial plan - 3-year projected P&L, monthly cash flow for year one, break-even, startup costs and sources of funds; for loans, debt service coverage (cash available for debt service / debt payments; lenders commonly look for 1.25x or more, but ask the lender).
- Risks and mitigations.
- Appendix - resumes, quotes, leases, letters of intent, permits.
Evidence rules
- Every market or demand claim cites a source (survey, interview count, competitor pricing page, government statistics) with the date accessed. Web lookups need network access; if it is off, mark claims as “to verify” rather than guessing figures.
- Prefer primary evidence: 15 customer interviews beat one analyst quote.
- Distinguish facts, estimates and targets with explicit wording (“we estimate”, “our target is”).
- No fabricated testimonials, letters or partnerships. Placeholders must be obvious:
[LETTER OF INTENT FROM <NAME> - TO OBTAIN].
Consistency checks
- Price x volume in the marketing section equals revenue in the financials
- Headcount in operations matches payroll in the model
- Startup costs in the funding request sum to the uses-of-funds table
- Break-even month in the summary matches the cash flow
- Milestone dates are achievable given the hiring plan and cash
- Executive summary contains no number that is absent from the body
Gotchas
- Writing 30 pages before testing whether anyone will pay. Ship the one-pager and the first assumption test, then expand.
- “If we capture 1% of the market” is not a sales plan; show the channel maths instead.
- Owner salary left out of costs makes a business look profitable that cannot pay its founder. Include it, even if deferred.
- Seasonality: monthly cash flow with flat revenue hides the months that need a credit line. Ask whether sales vary by season.
- Lenders and programmes reject plans that ignore their template. Use their headings verbatim.
Output
Deliver the plan in the requested format (markdown by default; .docx via the docx-documents skill if asked), plus:
## Open questions before this plan is final
- <data still needed, with owner>
- <assumptions untested>
- <sections needing professional review: financials (accountant), legal structure (lawyer)>What the zip contains
SKILL.md plus the interface file below. Nothing runs on install.
agents/openai.yaml
Gives the Codex app a display name, a one-line description and a starter prompt for this skill.
interface:
display_name: "Business Plan"
short_description: "A plan short enough to update monthly"
default_prompt: "$business-plan Turn my notes about a mobile dog grooming business into a one-page plan with the riskiest assumptions"About this skill
Written and maintained by the Codex Skills editors to the same editorial standard as every skill here: a selective description, a real workflow, known pitfalls and a way to verify the work. First published , last updated . MIT licensed. Something wrong or out of date? Tell us.