Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Render Mermaid diagrams as SVG and PNG using the Beautiful Mermaid library. Use when the user asks to render a Mermaid diagram.
.claude/skills/mkurman-beautiful-mermaid/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -53% | 0% |
----|------------|----------| | default | Light grey | General use | | dracula | Dark purple | Dark mode preference | | tokyo-night | Dark blue | Modern dark aesthetic | | tokyo-night-storm | Darker blue | Higher contrast | | nord | Dark arctic | Muted, calm visuals | | nord-light | Light arctic | Light mode with soft tones | | github-dark | GitHub dark | Matches GitHub UI | | github-light | GitHub light | Matches GitHub UI | | catppuccin-latte | Warm light | Soft pastel aesthetic | | solarized | Tan/cream | Solarized colour scheme | | one-dark | Atom dark | Atom editor aesthetic | | zinc-dark | Neutral dark | Minimal, no colour bias |
Check the render script output for the bg and fg values, or inspect the SVG's opening tag for --bg and --fg CSS custom properties.
-->|label| pipe notation, not -- label -->graph TD, graph LR, etc.)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,079 | 9,377 | -38% | 1 | 1 | 0% | 2,490 | 1,813 | -27% | 0 | 0 | — |
case-02 | fail→fail | 17,066 | 7,854 | -54% | 1 | 1 | 0% | 2,914 | 1,548 | -47% | 0 | 0 | — |
case-03 | fail→pass | 12,860 | 11,428 | -11% | 1 | 1 | 0% | 2,185 | 2,211 | +1% | 0 | 0 | — |
case-04 | fail→pass | 8,301 | 1,989 | -76% | 1 | 1 | 0% | 1,318 | 603 | -54% | 0 | 0 | — |
case-05 | fail→pass | 3,363 | 1,793 | -47% | 1 | 1 | 0% | 516 | 585 | +13% | 0 | 0 | — |
case-06 | fail→pass | 5,769 | 1,606 | -72% | 1 | 1 | 0% | 1,015 | 481 | -53% | 0 | 0 | — |
case-07 | fail→pass | 9,271 | 1,760 | -81% | 1 | 1 | 0% | 1,526 | 628 | -59% | 0 | 0 | — |
case-08 | fail→pass | 12,242 | 2,708 | -78% | 1 | 1 | 0% | 1,930 | 793 | -59% | 0 | 0 | — |
case-09 | fail→pass | 5,339 | 2,038 | -62% | 1 | 1 | 0% | 765 | 603 | -21% | 0 | 0 | — |
case-10 | fail→pass | 11,645 | 1,906 | -84% | 1 | 1 | 0% | 1,921 | 641 | -67% | 0 | 0 | — |
case-11 | fail→pass | 17,563 | 2,636 | -85% | 1 | 1 | 0% | 2,753 | 720 | -74% | 0 | 0 | — |
case-12 | pass→pass | 5,794 | 1,733 | -70% | 1 | 1 | 0% | 871 | 557 | -36% | 0 | 0 | — |
case-13 | pass→pass | 6,447 | 3,669 | -43% | 1 | 1 | 0% | 1,132 | 888 | -22% | 0 | 0 | — |
case-14 | pass→pass | 7,492 | 5,149 | -31% | 1 | 1 | 0% | 1,150 | 1,186 | +3% | 0 | 0 | — |
case-15 | fail→pass | 13,110 | 11,177 | -15% | 1 | 1 | 0% | 2,094 | 2,091 | -0% | 0 | 0 | — |
case-16 | fail→fail | 10,760 | 7,771 | -28% | 1 | 1 | 0% | 1,649 | 1,431 | -13% | 0 | 0 | — |
case-17 | pass→pass | 7,782 | 1,935 | -75% | 1 | 1 | 0% | 1,112 | 583 | -48% | 0 | 0 | — |
case-18 | pass→pass | 12,455 | 2,369 | -81% | 1 | 1 | 0% | 1,897 | 678 | -64% | 0 | 0 | — |
case-19 | pass→pass | 10,113 | 2,969 | -71% | 1 | 1 | 0% | 1,526 | 758 | -50% | 0 | 0 | — |
case-20 | pass→pass | 2,439 | 3,191 | +31% | 1 | 1 | 0% | 365 | 826 | +126% | 0 | 0 | — |
case-21 | pass→pass | 2,768 | 2,354 | -15% | 1 | 1 | 0% | 434 | 654 | +51% | 0 | 0 | — |
case-22 | pass→pass | 3,864 | 2,893 | -25% | 1 | 1 | 0% | 510 | 834 | +64% | 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 +50 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.