Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Extract TikZ diagrams from Beamer source, compile to PDF, convert to SVG with 0-based indexing. Use when updating TikZ diagrams for Quarto slides.
.claude/skills/brycewang-stanford-extract-tikz/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -68% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -25% | 0% |
Extract TikZ diagrams from the Beamer source, compile to multi-page PDF, and convert each page to SVG for use in Quarto slides.
Before compiling, verify that extract_tikz.tex matches the current Beamer source.
ls Slides/$ARGUMENTS*.tex\begin{tikzpicture} blocks from BeamerFigures/$ARGUMENTS/extract_tikz.texbashcd Figures/$ARGUMENTS
bashTEXINPUTS=../../Preambles:$TEXINPUTS xelatex -interaction=nonstopmode extract_tikz.tex
bashpdfinfo extract_tikz.pdf | grep "Pages:"
CRITICAL: PDF pages are 1-indexed, but output SVG files are 0-indexed!
bashPAGES=$(pdfinfo extract_tikz.pdf | grep "Pages:" | awk '{print $2}') for i in $(seq 1 $PAGES); do idx=$(printf "%02d" $((i-1))) pdf2svg extract_tikz.pdf tikz_exact_$idx.svg $i done
bashcd ../.. ./scripts/sync_to_docs.sh $ARGUMENTS
TikZ diagrams MUST be edited in the Beamer .tex file first, then copied verbatim to extract_tikz.tex. See .claude/rules/single-source-of-truth.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 2,765 | 5,528 | +100% | 1 | 1 | 0% | 349 | 742 | +113% | 0 | 0 | — |
case-07 | fail→pass | 8,036 | 4,282 | -47% | 1 | 1 | 0% | 1,421 | 1,392 | -2% | 0 | 0 | — |
case-08 | fail→pass | 7,651 | 5,377 | -30% | 1 | 1 | 0% | 1,322 | 1,409 | +7% | 0 | 0 | — |
case-02 | fail→fail | 4,203 | 5,744 | +37% | 1 | 1 | 0% | 200 | 967 | +384% | 0 | 0 | — |
case-03 | fail→fail | 4,737 | 4,895 | +3% | 1 | 1 | 0% | 329 | 838 | +155% | 0 | 0 | — |
case-04 | pass→pass | 10,189 | 4,407 | -57% | 1 | 1 | 0% | 1,890 | 1,319 | -30% | 0 | 0 | — |
case-05 | pass→pass | 4,475 | 5,585 | +25% | 1 | 1 | 0% | 886 | 1,619 | +83% | 0 | 0 | — |
case-06 | pass→pass | 5,137 | 4,141 | -19% | 1 | 1 | 0% | 932 | 1,278 | +37% | 0 | 0 | — |
case-09 | pass→pass | 3,033 | 1,813 | -40% | 1 | 1 | 0% | 537 | 839 | +56% | 0 | 0 | — |
case-10 | fail→pass | 14,069 | 2,078 | -85% | 1 | 1 | 0% | 2,699 | 856 | -68% | 0 | 0 | — |
case-11 | pass→pass | 6,974 | 3,587 | -49% | 1 | 1 | 0% | 1,280 | 1,245 | -3% | 0 | 0 | — |
case-12 | pass→pass | 9,798 | 2,520 | -74% | 1 | 1 | 0% | 1,865 | 1,027 | -45% | 0 | 0 | — |
case-13 | fail→fail | 9,283 | 1,901 | -80% | 1 | 1 | 0% | 1,720 | 810 | -53% | 0 | 0 | — |
case-14 | pass→pass | 6,357 | 3,842 | -40% | 1 | 1 | 0% | 1,110 | 1,310 | +18% | 0 | 0 | — |
case-15 | fail→pass | 9,321 | 1,933 | -79% | 1 | 1 | 0% | 1,680 | 872 | -48% | 0 | 0 | — |
case-16 | fail→pass | 12,318 | 6,146 | -50% | 1 | 1 | 0% | 2,205 | 1,655 | -25% | 0 | 0 | — |
case-17 | pass→pass | 8,546 | 4,274 | -50% | 1 | 1 | 0% | 1,491 | 1,292 | -13% | 0 | 0 | — |
case-18 | pass→pass | 10,688 | 2,722 | -75% | 1 | 1 | 0% | 2,007 | 1,077 | -46% | 0 | 0 | — |
case-19 | pass→pass | 9,571 | 1,551 | -84% | 1 | 1 | 0% | 1,782 | 734 | -59% | 0 | 0 | — |
case-20 | pass→pass | 12,045 | 5,850 | -51% | 1 | 1 | 0% | 2,207 | 1,516 | -31% | 0 | 0 | — |
case-21 | pass→pass | 3,033 | 1,968 | -35% | 1 | 1 | 0% | 559 | 852 | +52% | 0 | 0 | — |
case-22 | pass→pass | 11,138 | 4,670 | -58% | 1 | 1 | 0% | 2,006 | 1,515 | -24% | 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 19 counted toward the lift figure. The other 3 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 +23 percentage points is the difference between those two pass rates over the 19 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.