Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Categorize invoices and receipts by vendor, expense, and tax bucket, detect duplicates, and produce a tax-ready monthly summary. Use during bookkeeping, tax prep, or expense reimbursement.
.claude/skills/borghei-invoice-organizer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 172% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 117% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -11% | 0% |
Bulk-categorize a CSV of invoices or receipts, detect duplicates, and produce a tax-ready monthly summary.
invoice, invoices, receipt, receipts, expense, expenses, bookkeeping, accounting, tax, tax prep, categorization, vendor, reimbursement, monthly summary
Before categorizing, confirm these inputs. If any is unknown or vague, ASK — do not assume:
category_rules.json and cut the uncategorized bucketStop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
date,vendor,description,amount,currencybash python scripts/invoice_categorizer.py receipts.csv
Goal: Convert a month of unstructured receipts into a categorized, tax-ready summary in under 10 minutes.
Steps:
python scripts/invoice_categorizer.py receipts.csvassets/category_rules.json for any recurring vendorsassets/monthly_summary_template.mdExpected Output: Categorized expense list + monthly totals by category + duplicate-suspect list.
Time Estimate: 10 minutes/month after initial rules are seeded.
Goal: Catch double-entered receipts before they reach the books.
Steps:
python scripts/invoice_categorizer.py receipts.csv --jsonduplicates_suspected listExpected Output: Cleaned CSV with no duplicate rows.
Time Estimate: 2-3 minutes per month.
Goal: Find spend creep — vendors whose monthly total grew significantly without you noticing.
Steps:
Expected Output: Vendor-spend trend list with flagged growth.
Time Estimate: 15 minutes per quarter.
Reads a CSV of receipts/invoices and:
assets/category_rules.json (extensible)bash# Human-readable summary python scripts/invoice_categorizer.py receipts.csv # JSON for programmatic use python scripts/invoice_categorizer.py receipts.csv --json # Use a custom rules file python scripts/invoice_categorizer.py receipts.csv --rules my-rules.json
Expected CSV columns: date, vendor, description, amount (currency optional) Date formats accepted: YYYY-MM-DD, MM/DD/YYYY, DD/MM/YYYY
references/expense_categorization_guide.md — Standard expense categories, common tax buckets (US Schedule C, UK self-employment, generic), how to map vendors to categoriesassets/category_rules.json — Default rules; extend with your recurring vendorsassets/monthly_summary_template.md — Format for handing the monthly summary to an accountantfinance/ skills for budgeting and forecastingc-level-advisor/cs-cfo-advisor cash-flow workflows| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→pass | 4,095 | 5,183 | +27% | 1 | 1 | 0% | 853 | 2,321 | +172% | 0 | 0 | — |
case-01 | fail→fail | 6,931 | 13,588 | +96% | 1 | 1 | 0% | 1,315 | 3,619 | +175% | 0 | 0 | — |
case-02 | fail→fail | 4,501 | 5,792 | +29% | 1 | 1 | 0% | 728 | 2,232 | +207% | 0 | 0 | — |
case-04 | fail→pass | 6,308 | 6,741 | +7% | 1 | 1 | 0% | 1,236 | 2,686 | +117% | 0 | 0 | — |
case-05 | fail→pass | 13,179 | 6,298 | -52% | 1 | 1 | 0% | 2,024 | 2,316 | +14% | 0 | 0 | — |
case-06 | pass→pass | 9,031 | 1,631 | -82% | 1 | 1 | 0% | 1,377 | 1,595 | +16% | 0 | 0 | — |
case-07 | pass→pass | 5,784 | 1,633 | -72% | 1 | 1 | 0% | 922 | 1,589 | +72% | 0 | 0 | — |
case-08 | pass→pass | 6,938 | 2,387 | -66% | 1 | 1 | 0% | 1,229 | 1,754 | +43% | 0 | 0 | — |
case-09 | fail→pass | 9,987 | 3,551 | -64% | 1 | 1 | 0% | 1,650 | 1,919 | +16% | 0 | 0 | — |
case-10 | fail→pass | 11,322 | 3,104 | -73% | 1 | 1 | 0% | 1,994 | 1,779 | -11% | 0 | 0 | — |
case-11 | fail→pass | 8,517 | 1,930 | -77% | 1 | 1 | 0% | 1,398 | 1,637 | +17% | 0 | 0 | — |
case-12 | fail→pass | 17,262 | 1,733 | -90% | 1 | 1 | 0% | 2,901 | 1,561 | -46% | 0 | 0 | — |
case-13 | pass→pass | 9,758 | 7,784 | -20% | 1 | 1 | 0% | 1,380 | 2,414 | +75% | 0 | 0 | — |
case-14 | fail→pass | 10,074 | 1,625 | -84% | 1 | 1 | 0% | 1,530 | 1,577 | +3% | 0 | 0 | — |
case-15 | pass→pass | 11,768 | 9,367 | -20% | 1 | 1 | 0% | 1,751 | 2,789 | +59% | 0 | 0 | — |
case-16 | pass→pass | 13,262 | 6,036 | -54% | 1 | 1 | 0% | 1,839 | 2,166 | +18% | 0 | 0 | — |
case-17 | fail→pass | 5,581 | 1,659 | -70% | 1 | 1 | 0% | 772 | 1,519 | +97% | 0 | 0 | — |
case-18 | fail→fail | 12,563 | 2,511 | -80% | 1 | 1 | 0% | 1,841 | 1,716 | -7% | 0 | 0 | — |
case-19 | pass→pass | 9,666 | 7,971 | -18% | 1 | 1 | 0% | 1,560 | 2,421 | +55% | 0 | 0 | — |
case-20 | pass→pass | 7,804 | 5,958 | -24% | 1 | 1 | 0% | 1,284 | 2,236 | +74% | 0 | 0 | — |
case-21 | pass→pass | 11,320 | 7,405 | -35% | 1 | 1 | 0% | 1,751 | 2,479 | +42% | 0 | 0 | — |
case-22 | pass→pass | 11,308 | 6,724 | -41% | 1 | 1 | 0% | 1,766 | 2,258 | +28% | 0 | 0 | — |
case-23 | fail→pass | 11,891 | 2,328 | -80% | 1 | 1 | 0% | 1,890 | 1,633 | -14% | 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. 23 cases were attempted. The headline lift of +43 percentage points is the difference between those two pass rates over the 23 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.