Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build the tax habit self-employment requires — the setaside percentage from day one, the quarterly calendar, the records that make filing boring, and the no-withholding mindset shift nobody explains. Use when asked how do taxes work for my side income, how much should I set aside, what are estimated quarterly payments, or set up my freelance tax system. Produces the setaside rule with its honest range, the quarterly rhythm calendar (jurisdiction-flagged), the five-minute-a-week records system, a
.claude/skills/mohitagw15856-quarterly-tax-rhythm/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 39% | 0% |
Employment hides taxes inside withholding; self-employment hands you the gross and a delayed bill — and the first-year story is always the same: the money felt like income, got spent like income, and April arrived like a mugging. The fix isn't tax expertise; it's a rhythm: a fixed percentage siphoned to a separate account the day money lands, dated quarterly check-ins (most jurisdictions with estimated-payment systems run roughly quarterly — dates and rules are local, flagged throughout), and a records habit small enough to actually survive. This skill installs the rhythm and routes every actual number to a local professional, because rates and rules are jurisdiction-specific and this skill's job is that the money exists when the professional names the number.
Ask for these if not provided:
The % with its safety-margin logic · the on-receipt transfer rule · the account named · the one-way-glass rule verbatim]
The rhythm dates (typed, verify-locally for the real ones) · each check-in's 30-minute agenda · filing season's handoff date]
The sheet's columns · the weekly five minutes · the receipts folder · the deduction tags (typed, professional-verified)]
What the accountant gets: the export, the receipts, the questions list — and the first-meeting agenda if none is engaged yet]
> Rates, estimated-payment rules, deadlines, and deductibility are jurisdiction-specific — the percentages here are safety margins and the calendar is a rhythm; a local tax professional supplies the real numbers, and this system's job is making their work (and bill) small. Not tax advice.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 30,537 | 35,828 | +17% | 1 | 1 | 0% | 3,630 | 5,602 | +54% | 0 | 0 | — |
case-02 | fail→fail | 45,867 | 27,639 | -40% | 1 | 1 | 0% | 6,679 | 4,469 | -33% | 0 | 0 | — |
case-03 | fail→pass | 47,954 | 36,423 | -24% | 1 | 1 | 0% | 6,876 | 5,509 | -20% | 0 | 0 | — |
case-04 | pass→pass | 20,691 | 25,195 | +22% | 1 | 1 | 0% | 2,231 | 3,993 | +79% | 0 | 0 | — |
case-05 | pass→pass | 19,921 | 23,891 | +20% | 1 | 1 | 0% | 2,381 | 4,327 | +82% | 0 | 0 | — |
case-06 | fail→pass | 21,756 | 19,623 | -10% | 1 | 1 | 0% | 2,271 | 3,157 | +39% | 0 | 0 | — |
case-07 | fail→pass | 33,306 | 22,802 | -32% | 1 | 1 | 0% | 2,752 | 4,299 | +56% | 0 | 0 | — |
case-08 | fail→pass | 18,463 | 18,883 | +2% | 1 | 1 | 0% | 2,069 | 3,780 | +83% | 0 | 0 | — |
case-09 | pass→pass | 16,221 | 27,200 | +68% | 1 | 1 | 0% | 1,943 | 4,241 | +118% | 0 | 0 | — |
case-10 | pass→pass | 16,145 | 29,214 | +81% | 1 | 1 | 0% | 2,324 | 4,442 | +91% | 0 | 0 | — |
case-11 | fail→pass | 18,754 | 24,722 | +32% | 1 | 1 | 0% | 3,109 | 4,307 | +39% | 0 | 0 | — |
case-12 | fail→fail | 15,935 | 17,898 | +12% | 1 | 1 | 0% | 2,100 | 3,561 | +70% | 0 | 0 | — |
case-13 | fail→fail | 13,785 | 20,191 | +46% | 1 | 1 | 0% | 1,958 | 3,734 | +91% | 0 | 0 | — |
case-14 | pass→pass | 12,606 | 13,159 | +4% | 1 | 1 | 0% | 1,996 | 3,283 | +64% | 0 | 0 | — |
case-19 | pass→pass | 10,433 | 20,062 | +92% | 1 | 1 | 0% | 1,761 | 3,990 | +127% | 0 | 0 | — |
case-15 | fail→pass | 13,649 | 16,954 | +24% | 1 | 1 | 0% | 1,853 | 3,421 | +85% | 0 | 0 | — |
case-16 | fail→fail | 19,699 | 18,109 | -8% | 1 | 1 | 0% | 2,514 | 4,073 | +62% | 0 | 0 | — |
case-17 | fail→fail | 18,388 | 14,989 | -18% | 1 | 1 | 0% | 2,590 | 3,707 | +43% | 0 | 0 | — |
case-18 | fail→fail | 19,132 | 15,380 | -20% | 1 | 1 | 0% | 2,579 | 3,218 | +25% | 0 | 0 | — |
case-20 | fail→pass | 19,405 | 15,620 | -20% | 1 | 1 | 0% | 3,878 | 3,992 | +3% | 0 | 0 | — |
case-21 | fail→fail | 23,347 | 22,802 | -2% | 1 | 1 | 0% | 3,309 | 4,826 | +46% | 0 | 0 | — |
case-22 | pass→pass | 20,218 | 16,712 | -17% | 1 | 1 | 0% | 2,906 | 4,197 | +44% | 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 +32 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.