Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Find out which clients and jobs actually make money -- allocate real costs (labor hours, materials, subs, overhead) against what each job billed, and rank every client by true margin. Most service businesses discover their biggest client is their least profitable.
.claude/skills/onewave-ai-job-profitability-analyzer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 194% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 112% | 0% |
| case-13 | ✓→✗ | ▼ Worse | 47% | 0% |
| case-17 | ✓→✗ | ▼ Worse | 31% | 0% |
Revenue is vanity, margin per job is survival. Service businesses (agencies, contractors, consultancies, trades) routinely lose money on their "best" clients because nobody allocates the hours. Input: invoices/revenue by job or client, time tracking exports, materials/sub costs, payroll rates, and a rough overhead number. Output: profit per job, per client, per service line -- with receipts.
profitability-report.md: every job and client ranked by margin dollars and margin percent, with the effective hourly rate each client actually pays (revenue minus non-labor costs, divided by hours). This one column reorders most people's client list.pricing-strategy for the repricing conversation.INCOMPLETE, not a guess -- and chronic time-tracking gaps get called out as the root problem.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,041 | 13,692 | +14% | 1 | 1 | 0% | 1,655 | 2,257 | +36% | 0 | 0 | — |
case-02 | fail→pass | 4,178 | 10,105 | +142% | 1 | 1 | 0% | 796 | 2,342 | +194% | 0 | 0 | — |
case-03 | pass→pass | 8,490 | 12,673 | +49% | 1 | 1 | 0% | 1,682 | 2,618 | +56% | 0 | 0 | — |
case-04 | pass→pass | 7,894 | 6,332 | -20% | 1 | 1 | 0% | 1,534 | 1,628 | +6% | 0 | 0 | — |
case-05 | pass→pass | 6,085 | 6,786 | +12% | 1 | 1 | 0% | 1,070 | 1,735 | +62% | 0 | 0 | — |
case-06 | pass→pass | 13,925 | 16,996 | +22% | 1 | 1 | 0% | 2,353 | 3,059 | +30% | 0 | 0 | — |
case-07 | fail→pass | 16,235 | 15,982 | -2% | 1 | 1 | 0% | 2,206 | 2,789 | +26% | 0 | 0 | — |
case-08 | pass→pass | 6,479 | 8,047 | +24% | 1 | 1 | 0% | 1,318 | 2,142 | +63% | 0 | 0 | — |
case-09 | pass→pass | 7,158 | 14,269 | +99% | 1 | 1 | 0% | 1,380 | 2,856 | +107% | 0 | 0 | — |
case-10 | fail→pass | 11,341 | 22,221 | +96% | 1 | 1 | 0% | 2,237 | 4,747 | +112% | 0 | 0 | — |
case-11 | pass→pass | 11,009 | 14,095 | +28% | 1 | 1 | 0% | 1,732 | 2,663 | +54% | 0 | 0 | — |
case-12 | pass→pass | 10,696 | 13,339 | +25% | 1 | 1 | 0% | 2,053 | 3,095 | +51% | 0 | 0 | — |
case-13 | pass→fail | 11,744 | 14,742 | +26% | 1 | 1 | 0% | 2,525 | 3,719 | +47% | 0 | 0 | — |
case-14 | pass→pass | 10,998 | 9,925 | -10% | 1 | 1 | 0% | 1,620 | 2,587 | +60% | 0 | 0 | — |
case-15 | pass→pass | 10,961 | 13,713 | +25% | 1 | 1 | 0% | 1,811 | 2,917 | +61% | 0 | 0 | — |
case-16 | pass→pass | 9,127 | 10,349 | +13% | 1 | 1 | 0% | 1,655 | 2,265 | +37% | 0 | 0 | — |
case-17 | pass→fail | 12,677 | 13,252 | +5% | 1 | 1 | 0% | 2,127 | 2,787 | +31% | 0 | 0 | — |
case-18 | pass→pass | 8,653 | 6,263 | -28% | 1 | 1 | 0% | 1,747 | 1,825 | +4% | 0 | 0 | — |
case-19 | pass→pass | 10,440 | 6,956 | -33% | 1 | 1 | 0% | 1,949 | 1,801 | -8% | 0 | 0 | — |
case-20 | fail→fail | 5,971 | 4,755 | -20% | 1 | 1 | 0% | 1,079 | 1,349 | +25% | 0 | 0 | — |
case-21 | fail→fail | 12,629 | 10,471 | -17% | 1 | 1 | 0% | 2,170 | 2,370 | +9% | 0 | 0 | — |
case-22 | fail→fail | 10,602 | 11,688 | +10% | 1 | 1 | 0% | 2,064 | 2,593 | +26% | 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 +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
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.