Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Design an expense-tracking sheet that survives real receipts — the capture-at-spend habit, the category set that matches reimbursement or tax rules, the receipt-link discipline, and the month-end close that takes minutes because the work happened at spend-time. Use when asked track my business expenses, build an expense sheet for the team, get ready for reimbursement/tax season, or my shoebox of receipts needs a system. Produces the sheet structure, the capture ritual, the category mapping to th
.claude/skills/mohitagw15856-expense-sheet-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 29% | 0% |
Expense tracking has one law: it happens at spend-time or it happens badly. The March reconstruction — a shoebox of receipts, a statement, and archaeology — produces worse data at fifty times the cost of a fifteen-second capture habit. The sheet's design serves that habit: columns light enough to fill from a phone, categories that map to the downstream consumer's rules (the reimbursement policy, the tax return's lines — not aesthetic taxonomy), receipts linked not shoeboxed, and a month-end close that's minutes because every entry already exists.
Ask for these if not provided:
Columns with the ≤15-second test applied · the dropdown source · the receipt-link route]
| Sheet category (= downstream's) | Downstream line | Receipt required? (verify) | |---|---|---|
The 15-second flow, phone-first · subscriptions as auto-rows · the odd-one note rule]
Statement reconcile → chase gaps → subtotals in downstream format → export filed]
> Reimbursement rules and deduction categories are the downstream's law — policy documents and local tax professionals define them; this sheet transcribes, never decides. Not tax advice.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 11,709 | 13,098 | +12% | 1 | 1 | 0% | 2,102 | 3,313 | +58% | 0 | 0 | — |
case-01 | fail→pass | 18,204 | 15,502 | -15% | 1 | 1 | 0% | 3,094 | 3,644 | +18% | 0 | 0 | — |
case-02 | fail→pass | 19,243 | 15,290 | -21% | 1 | 1 | 0% | 3,139 | 3,670 | +17% | 0 | 0 | — |
case-03 | fail→fail | 18,487 | 13,890 | -25% | 1 | 1 | 0% | 3,170 | 3,246 | +2% | 0 | 0 | — |
case-05 | pass→pass | 5,612 | 5,817 | +4% | 1 | 1 | 0% | 1,040 | 2,078 | +100% | 0 | 0 | — |
case-06 | pass→pass | 5,099 | 5,986 | +17% | 1 | 1 | 0% | 939 | 2,178 | +132% | 0 | 0 | — |
case-07 | pass→pass | 9,487 | 8,908 | -6% | 1 | 1 | 0% | 1,443 | 2,580 | +79% | 0 | 0 | — |
case-08 | pass→pass | 12,276 | 8,565 | -30% | 1 | 1 | 0% | 1,966 | 2,550 | +30% | 0 | 0 | — |
case-09 | pass→pass | 12,579 | 9,792 | -22% | 1 | 1 | 0% | 1,948 | 2,439 | +25% | 0 | 0 | — |
case-10 | pass→pass | 13,931 | 15,310 | +10% | 1 | 1 | 0% | 2,155 | 3,544 | +64% | 0 | 0 | — |
case-11 | pass→pass | 12,179 | 13,563 | +11% | 1 | 1 | 0% | 1,885 | 3,391 | +80% | 0 | 0 | — |
case-12 | pass→pass | 6,149 | 2,168 | -65% | 1 | 1 | 0% | 934 | 1,512 | +62% | 0 | 0 | — |
case-22 | pass→pass | 12,805 | 7,345 | -43% | 1 | 1 | 0% | 2,159 | 2,270 | +5% | 0 | 0 | — |
case-13 | pass→pass | 11,668 | 10,762 | -8% | 1 | 1 | 0% | 1,714 | 2,701 | +58% | 0 | 0 | — |
case-14 | pass→pass | 14,081 | 11,278 | -20% | 1 | 1 | 0% | 2,134 | 2,718 | +27% | 0 | 0 | — |
case-15 | pass→pass | 9,027 | 8,703 | -4% | 1 | 1 | 0% | 1,322 | 2,476 | +87% | 0 | 0 | — |
case-16 | pass→pass | 13,607 | 14,702 | +8% | 1 | 1 | 0% | 2,144 | 3,307 | +54% | 0 | 0 | — |
case-17 | fail→pass | 13,086 | 6,539 | -50% | 1 | 1 | 0% | 2,129 | 2,188 | +3% | 0 | 0 | — |
case-18 | pass→pass | 11,684 | 14,482 | +24% | 1 | 1 | 0% | 1,834 | 3,144 | +71% | 0 | 0 | — |
case-19 | fail→pass | 11,540 | 8,474 | -27% | 1 | 1 | 0% | 1,677 | 2,368 | +41% | 0 | 0 | — |
case-20 | pass→pass | 8,212 | 2,885 | -65% | 1 | 1 | 0% | 1,376 | 1,552 | +13% | 0 | 0 | — |
case-21 | fail→pass | 9,855 | 5,600 | -43% | 1 | 1 | 0% | 1,517 | 1,950 | +29% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +23 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.