Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Convert a Markdown file to PDF with GitHub-style formatting using the md2pdf tool.
.claude/skills/joshukraine-md2pdf/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -36% | 0% |
Convert Markdown files to PDF with GitHub-style formatting using the md2pdf tool.
$ARGUMENTS: File paths or glob patterns to convert (e.g., docs/guide.md, docs/*.md)-o FILE: Custom output path (only valid with a single input file)$ARGUMENTS are provided, use them as the file pathsmd2pdf with the specified filesThe md2pdf tool uses pandoc + weasyprint with GitHub-style CSS. The CSS and script live in dotfiles (~/dotfiles/bin/.local/bin/). The MD2PDF_CSS environment variable can override the default CSS path.
bash# Convert a single file /md2pdf docs/guide.md # Convert multiple files with a glob /md2pdf docs/qa-handoffs/qa-guide-*.md # Convert with a custom output name /md2pdf -o custom-name.pdf docs/guide.md
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | fail→fail | 4,550 | 8,173 | +80% | 1 | 1 | 0% | 493 | 596 | +21% | 0 | 0 | — |
case-07 | fail→fail | 7,801 | 5,797 | -26% | 1 | 1 | 0% | 1,399 | 546 | -61% | 0 | 0 | — |
case-01 | fail→fail | 8,578 | 7,669 | -11% | 1 | 1 | 0% | 1,373 | 581 | -58% | 0 | 0 | — |
case-02 | fail→fail | 9,054 | 5,516 | -39% | 1 | 1 | 0% | 1,573 | 568 | -64% | 0 | 0 | — |
case-03 | fail→fail | 3,558 | 4,481 | +26% | 1 | 1 | 0% | 475 | 404 | -15% | 0 | 0 | — |
case-04 | fail→pass | 6,380 | 2,980 | -53% | 1 | 1 | 0% | 972 | 718 | -26% | 0 | 0 | — |
case-05 | fail→fail | 5,191 | 5,934 | +14% | 1 | 1 | 0% | 885 | 653 | -26% | 0 | 0 | — |
case-06 | fail→pass | 8,191 | 2,257 | -72% | 1 | 1 | 0% | 1,333 | 666 | -50% | 0 | 0 | — |
case-09 | fail→fail | 30,392 | 5,118 | -83% | 1 | 1 | 0% | 6,156 | 430 | -93% | 0 | 0 | — |
case-10 | fail→pass | 4,520 | 8,433 | +87% | 1 | 1 | 0% | 605 | 715 | +18% | 0 | 0 | — |
case-11 | fail→fail | 5,616 | 7,955 | +42% | 1 | 1 | 0% | 831 | 905 | +9% | 0 | 0 | — |
case-12 | fail→fail | 9,766 | 5,023 | -49% | 1 | 1 | 0% | 1,871 | 464 | -75% | 0 | 0 | — |
case-13 | fail→fail | 3,736 | 5,991 | +60% | 1 | 1 | 0% | 359 | 516 | +44% | 0 | 0 | — |
case-14 | fail→pass | 6,326 | 1,742 | -72% | 1 | 1 | 0% | 1,138 | 561 | -51% | 0 | 0 | — |
case-15 | fail→pass | 4,290 | 2,357 | -45% | 1 | 1 | 0% | 683 | 435 | -36% | 0 | 0 | — |
case-16 | fail→fail | 9,196 | 5,732 | -38% | 1 | 1 | 0% | 1,651 | 508 | -69% | 0 | 0 | — |
case-17 | fail→pass | 13,940 | 8,813 | -37% | 1 | 1 | 0% | 2,380 | 1,200 | -50% | 0 | 0 | — |
case-18 | fail→pass | 3,851 | 16,156 | +320% | 1 | 1 | 0% | 618 | 2,515 | +307% | 0 | 0 | — |
case-19 | fail→pass | 2,916 | 2,068 | -29% | 1 | 1 | 0% | 383 | 568 | +48% | 0 | 0 | — |
case-20 | pass→pass | 9,113 | 5,551 | -39% | 1 | 1 | 0% | 1,454 | 1,117 | -23% | 0 | 0 | — |
case-21 | fail→fail | 2,691 | 11,025 | +310% | 1 | 1 | 0% | 349 | 771 | +121% | 0 | 0 | — |
case-22 | pass→pass | 11,448 | 8,116 | -29% | 1 | 1 | 0% | 2,132 | 1,501 | -30% | 0 | 0 | — |
case-23 | pass→pass | 8,658 | 5,287 | -39% | 1 | 1 | 0% | 1,541 | 1,204 | -22% | 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, and 12 counted toward the lift figure. The other 11 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 +35 percentage points is the difference between those two pass rates over the 12 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.