Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.
.claude/skills/mkurman-pdf/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-09 | ✓→✓ | ＝ Same ✓ | -15% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -70% | 0% |
|-----------|--------------| | Merge PDFs | pypdf | writer.add_page(page) | | Split PDFs | pypdf | One page per file | | Extract text | pdfplumber | page.extract_text() | | Extract tables | pdfplumber | page.extract_tables() | | Create PDFs | reportlab | Canvas or Platypus | | Command line merge | qpdf | qpdf --empty --pages ... | | OCR scanned PDFs | pytesseract | Convert to image first | | Fill PDF forms | pdf-lib or pypdf (see FORMS.md) | See FORMS.md |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 11,405 | 7,876 | -31% | 1 | 1 | 0% | 1,942 | 1,655 | -15% | 0 | 0 | — |
case-01 | fail→pass | 7,475 | 2,914 | -61% | 1 | 1 | 0% | 1,274 | 636 | -50% | 0 | 0 | — |
case-02 | fail→pass | 7,374 | 6,893 | -7% | 1 | 1 | 0% | 1,312 | 1,536 | +17% | 0 | 0 | — |
case-03 | pass→pass | 11,742 | 3,064 | -74% | 1 | 1 | 0% | 2,216 | 665 | -70% | 0 | 0 | — |
case-08 | pass→pass | 13,495 | 7,191 | -47% | 1 | 1 | 0% | 2,349 | 1,411 | -40% | 0 | 0 | — |
case-04 | pass→pass | 12,862 | 7,821 | -39% | 1 | 1 | 0% | 2,278 | 1,679 | -26% | 0 | 0 | — |
case-05 | pass→pass | 8,381 | 4,060 | -52% | 1 | 1 | 0% | 1,539 | 942 | -39% | 0 | 0 | — |
case-06 | pass→pass | 11,477 | 8,434 | -27% | 1 | 1 | 0% | 1,950 | 1,744 | -11% | 0 | 0 | — |
case-07 | pass→pass | 11,648 | 2,907 | -75% | 1 | 1 | 0% | 1,861 | 726 | -61% | 0 | 0 | — |
case-10 | pass→pass | 9,597 | 7,692 | -20% | 1 | 1 | 0% | 1,645 | 1,568 | -5% | 0 | 0 | — |
case-11 | pass→pass | 6,186 | 4,684 | -24% | 1 | 1 | 0% | 1,109 | 1,021 | -8% | 0 | 0 | — |
case-12 | pass→pass | 5,198 | 3,141 | -40% | 1 | 1 | 0% | 879 | 666 | -24% | 0 | 0 | — |
case-13 | pass→pass | 12,883 | 10,675 | -17% | 1 | 1 | 0% | 2,140 | 2,098 | -2% | 0 | 0 | — |
case-14 | pass→pass | 11,945 | 5,421 | -55% | 1 | 1 | 0% | 1,948 | 1,114 | -43% | 0 | 0 | — |
case-15 | pass→pass | 4,729 | 1,655 | -65% | 1 | 1 | 0% | 694 | 442 | -36% | 0 | 0 | — |
case-16 | pass→pass | 4,553 | 2,162 | -53% | 1 | 1 | 0% | 762 | 522 | -31% | 0 | 0 | — |
case-17 | pass→pass | 7,032 | 2,462 | -65% | 1 | 1 | 0% | 1,075 | 606 | -44% | 0 | 0 | — |
case-18 | pass→pass | 9,511 | 3,757 | -60% | 1 | 1 | 0% | 1,781 | 839 | -53% | 0 | 0 | — |
case-19 | pass→pass | 6,856 | 4,106 | -40% | 1 | 1 | 0% | 1,193 | 838 | -30% | 0 | 0 | — |
case-20 | fail→pass | 30,915 | 11,196 | -64% | 1 | 1 | 0% | 1,240 | 1,999 | +61% | 0 | 0 | — |
case-21 | pass→pass | 9,787 | 7,478 | -24% | 1 | 1 | 0% | 1,791 | 1,490 | -17% | 0 | 0 | — |
case-22 | pass→pass | 14,610 | 9,017 | -38% | 1 | 1 | 0% | 2,664 | 1,974 | -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, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +14 percentage points is the difference between those two pass rates over the 21 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.