Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Point Cowork at your invoices and payments records -- builds an AR aging report, drafts escalating follow-up emails matched to each invoice's age and the client relationship, and runs as a weekly scheduled task so nothing slips past 30 days unnoticed.
.claude/skills/onewave-ai-cowork-invoice-chaser/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 287% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 62% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 31% | 0% |
Chase receivables the way a firm-but-warm operator does: know exactly who owes what and for how long, escalate on a schedule, and never send a dunning email to someone who already paid. Inputs: a folder of issued invoices (PDF/.docx) and payment evidence (bank exports, remittance emails, a payments .csv, or a paid-list the user maintains).
key account, standard, problem); key accounts never get form-letter tone regardless of age.As a weekly Cowork scheduled task: re-scan the folders, reconcile new payments, advance escalation stages, and deliver the digest with drafts ready for approval. The run is idempotent -- no invoice is double-chased within its stage window.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→pass | 8,921 | 10,865 | +22% | 1 | 1 | 0% | 1,266 | 2,353 | +86% | 0 | 0 | — |
case-01 | fail→fail | 17,213 | 29,765 | +73% | 1 | 1 | 0% | 2,529 | 4,206 | +66% | 0 | 0 | — |
case-02 | fail→pass | 8,322 | 21,933 | +164% | 1 | 1 | 0% | 1,221 | 4,720 | +287% | 0 | 0 | — |
case-03 | pass→pass | 15,006 | 10,592 | -29% | 1 | 1 | 0% | 2,249 | 2,612 | +16% | 0 | 0 | — |
case-04 | fail→fail | 6,292 | 10,086 | +60% | 1 | 1 | 0% | 1,006 | 2,310 | +130% | 0 | 0 | — |
case-06 | fail→fail | 8,172 | 13,046 | +60% | 1 | 1 | 0% | 1,376 | 2,622 | +91% | 0 | 0 | — |
case-07 | fail→pass | 8,434 | 9,975 | +18% | 1 | 1 | 0% | 1,464 | 2,369 | +62% | 0 | 0 | — |
case-08 | fail→pass | 9,819 | 11,322 | +15% | 1 | 1 | 0% | 1,839 | 2,809 | +53% | 0 | 0 | — |
case-09 | fail→pass | 18,913 | 7,639 | -60% | 1 | 1 | 0% | 1,473 | 1,934 | +31% | 0 | 0 | — |
case-10 | fail→pass | 8,693 | 10,985 | +26% | 1 | 1 | 0% | 1,261 | 2,375 | +88% | 0 | 0 | — |
case-11 | fail→fail | 8,145 | 9,835 | +21% | 1 | 1 | 0% | 1,306 | 2,074 | +59% | 0 | 0 | — |
case-12 | fail→fail | 7,147 | 5,428 | -24% | 1 | 1 | 0% | 1,248 | 1,529 | +23% | 0 | 0 | — |
case-13 | fail→fail | 5,917 | 5,413 | -9% | 1 | 1 | 0% | 995 | 1,594 | +60% | 0 | 0 | — |
case-14 | fail→pass | 10,421 | 8,154 | -22% | 1 | 1 | 0% | 1,599 | 1,960 | +23% | 0 | 0 | — |
case-15 | fail→fail | 2,937 | 4,712 | +60% | 1 | 1 | 0% | 398 | 1,412 | +255% | 0 | 0 | — |
case-16 | fail→fail | 1,931 | 7,433 | +285% | 1 | 1 | 0% | 237 | 1,676 | +607% | 0 | 0 | — |
case-17 | pass→pass | 6,785 | 36,630 | +440% | 1 | 1 | 0% | 1,072 | 1,823 | +70% | 0 | 0 | — |
case-18 | fail→fail | 3,888 | 10,276 | +164% | 1 | 1 | 0% | 578 | 2,119 | +267% | 0 | 0 | — |
case-19 | fail→fail | 7,890 | 2,881 | -63% | 1 | 1 | 0% | 1,136 | 976 | -14% | 0 | 0 | — |
case-20 | pass→pass | 5,850 | 11,433 | +95% | 1 | 1 | 0% | 894 | 2,410 | +170% | 0 | 0 | — |
case-21 | pass→pass | 6,241 | 4,197 | -33% | 1 | 1 | 0% | 1,000 | 1,260 | +26% | 0 | 0 | — |
case-22 | pass→pass | 11,078 | 10,237 | -8% | 1 | 1 | 0% | 1,917 | 2,504 | +31% | 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.