cash-week

---
name: cash-week
description: Builds a 13 week rolling cash forecast from your real bank balance, your committed payments and the invoices you believe will be paid, then names the first week the balance goes negative.
---

# Know the week you run out of cash, before you get there

You give me today's bank balance, what is committed to go out, and which invoices you expect to be paid and when. You get back thirteen weekly columns, the closing balance for each one, and the date of the first week you go under.

## What it does

1. **Start from the cleared bank balance, not the accounts.** Ask for today's cleared balance on every business account, plus any agreed overdraft limit and the floor the owner will not go below. Add them into one opening figure and write down which accounts it came from. This is the direct method, which is what a 13 week forecast is: Atlar states that most 13 week forecasts "use the 'direct' forecasting method, tracking actual cash receipts and payments" rather than working down from a profit figure.
2. **Lay out thirteen weekly columns starting from this Monday, and label each one with its real date.** Thirteen weeks is one quarter, and Stampli's guide gives the reason to use it rather than a month or a year: "Thirteen weeks balances actionability and visibility: near weeks are accurate enough to manage payment timing, the full quarter is far enough to act on a shortfall."
3. **Enter committed money out from its own calendar, week by week, before anything else.** Payroll and the PAYE that follows it, rent, VAT, loan and finance repayments, insurance, every direct debit, and supplier invoices already approved. Stampli's structure puts already scheduled payments in weeks 1 to 2, approved invoices in weeks 2 to 5, unapproved invoices in weeks 3 to 6, recurring payments across all weeks, and open purchase orders from week 5 to 13. Use the owner's actual dates where they exist and that shape only where they do not.
4. **Enter money in as one line per invoice, dated the week the cash is genuinely expected to land, never the invoice due date.** Ask the owner, for each invoice: has this customer paid you late before, and by how long? Move the line by that many days.
5. **Ask the owner for a payment likelihood on every receipt line and never supply one yourself.** Certain, likely, or doubtful, in their words about their own customer. Anthropic's own cash-flow-snapshot skill defaults to a plus or minus 30 per cent band when a customer has fewer than three payments on record. This skill deliberately does not do that. An invoice the owner will not put a likelihood against is taken out of the forecast entirely and listed underneath in a section called "Not counted", with the amount shown, so the owner can see what has been left out.
6. **Keep work you have not yet won out of the receipts entirely.** Quotes, pipeline, "he always comes back in March" and anything else unsigned goes below the closing balance line as a separate scenario row. It never touches the closing balance the crunch week is read from.
7. **Work out net movement and closing balance for every week, then name the first week the closing balance goes below zero, and separately the first week it goes below the owner's stated floor.** Give both as a real date, for example "week commencing 12 October". If neither happens in thirteen weeks, say that in one sentence and give the lowest week and its balance.
8. **Put a confidence label on each block of weeks and say out loud that it is a target, not a measurement.** Float's guide cites commonly used targets of roughly 90 to 95 per cent accuracy in weeks 1 to 4, 85 to 90 per cent in weeks 5 to 8 and 70 to 85 per cent in weeks 9 to 13, and states plainly that "no published study measures 13-week forecast accuracy against realised cash for businesses of this size". Repeat that caveat in the output. Where a lender is involved, note that permitted variance in covenants commonly sits between 10 and 20 per cent, with 15 per cent recurring most often, tested over rolling four week periods rather than single weeks.
9. **Finish with the three named moves that change the crunch week, and the roll-forward instruction.** Each move is a specific invoice to chase, a specific payment to move, or a specific cost to stop, with the pounds and the week it lands in. Then tell the owner to come back in a week with actuals so the model can be rolled forward: Stampli's discipline is to "record actuals against last week's forecast, explain the variances, and roll the model forward", concentrating on the ten largest differences so the weekly update stays under an hour.

## Then it checks

1. The opening balance equals the bank balance the owner gave, to the penny, and the report names the accounts it was taken from.
2. Every receipt line carries a likelihood the owner supplied in their own words. Any line without one appears in the "Not counted" list and in no other total.
3. Payroll, rent, VAT and loan repayments appear in every week they are actually due. A thirteen week window contains at least three monthly payroll runs, so if fewer than three appear, a date has been missed.
4. The arithmetic closes: each week's closing balance equals the previous closing plus receipts minus payments, and week 13's closing equals the opening balance plus the sum of all thirteen net movements.
5. The crunch week is stated as a calendar date, or the output says clearly that no week goes negative and gives the lowest week instead.
6. No figure in the forecast is one the owner did not supply or that cannot be traced to an invoice, a bill or a bank line.

Any check fails: name it, redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only.
- Never invent a fact, a number or a quote.
- Never invent a payment probability, a payment date or an average collection period. If the owner has not given it, the line is excluded and listed as not counted. A forecast built on a guessed likelihood is worse than no forecast, because it gets believed.
- Never smooth, round up or move a payment to make a week survive. If the balance goes negative, show it going negative.
- This is a working document, not financial, tax or insolvency advice, and it is prepared for the owner's accountant to check before it is shown to a bank, a lender or an investor. If the forecast shows the business unable to pay its debts as they fall due, say so in plain words and tell the owner to speak to their accountant or a licensed insolvency practitioner now, not next quarter.

## Built from
- Anthropic, "Introducing Claude for Small Business", https://www.anthropic.com/news/claude-for-small-business, published 13 May 2026: the Business Pulse Dashboard workflow, which surfaces "your cash position through Intuit QuickBooks, sales trend, pipeline movement, this week's commitments" on one scheduled page. This skill takes the one page, on a schedule, cash first idea from it and narrows it to the single question an owner actually loses sleep over.
- anthropics/knowledge-work-plugins, small-business plugin, cash-flow-snapshot skill, https://github.com/anthropics/knowledge-work-plugins/blob/main/small-business/skills/cash-flow-snapshot/SKILL.md, 23,911 stars read from api.github.com: the named risk flag format ("Payroll hits April 15, low-band cash April 14, shortfall risk"), the confidence band idea, and its plus or minus 30 per cent default, which this skill refuses and replaces with a hard exclusion.
- Stampli, "What is a 13-week cash flow forecast, and why 13 weeks?", https://www.stampli.com/resources/13-week-cash-flow-forecast/, no publication date shown, read 7 September 2026: the reason for the thirteen week horizon, the week by week placement of scheduled, approved, unapproved and recurring payments, and the weekly roll-forward with the ten largest variances in under an hour.
- Float, "How Accurate Are 13-Week Cash Flow Forecasts?", https://www.floatapp.com/blog/13-week-cash-flow-forecast-accuracy, no publication date shown, read 7 September 2026: the 90 to 95, 85 to 90 and 70 to 85 per cent accuracy bands, the statement that these are targets rather than measurements, and the 10 to 20 per cent lender covenant variance tested over rolling four week periods.
- Atlar, "What is the 13-week cash flow forecast?", https://www.atlar.com/learn/what-is-the-13-week-cash-flow-forecast, no publication date shown, read 7 September 2026: the direct method definition, the three input structure of opening cash, receipts and disbursements, and the rule that accuracy degrades as the forecast range increases.