Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A humanities lecture: how Renaissance linear perspective reshaped early cartography — sources, argument, and evidence. Built as a decision-grade academic research deck for faculty, graduate seminar.
.claude/skills/nexu-io-html-ppt-zhangzara-vellum/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 19% | 0% |
> Deep navy canvas with warm-yellow italic Cormorant serifs and a single dusty teal accent. A quiet, scholarly aesthetic.
A single self-contained HTML deck — typography, palette, decorative system, and slide vocabulary are all tuned together. Mixing layouts across templates breaks the system; stay inside this one.
Anything that should feel scholarly, literary, and quietly intelligent: research synthesis, white papers, academic and policy briefs, advisory deliverables, longform editorial pieces, founder reflections. Equally strong for any deck — including tech, business, or creator work — that wants a calm, considered atmosphere instead of energetic visuals.
Contexts that need visual heat or pop — the navy + warm-yellow italic-Cormorant aesthetic is intentionally low-tempo.
example.html into the user's workspace as the working file.numbers, names, dates, and section labels. Match existing dimensions when swapping image placeholders.
restructure the layout grid, or strip decorative elements (corner brackets, paper grain, geometric shapes, illustrated SVGs). They are part of the identity.
than the demo holds, duplicate an existing slide of the most appropriate layout. If less, drop slides from the bottom. Update page-number labels.
doesn't have, design it from scratch using the same fonts, palette, decorative vocabulary, spacing rhythm, and component grammar — never bail to a different template.
assets/deck-stage.js or inline keyboard handler, leave it intact.
Emit between <artifact> tags:
<artifact identifier="zhangzara-vellum" type="text/html" title="Deck Title">
<!doctype html>
<html>...</html>
</artifact>Vendored from upstream MIT-licensed zarazhangrui/beautiful-html-templates.
The full upstream MIT license text — including the original copyright notice — ships in this skill at LICENSE and must be redistributed alongside any copy of example.html, template.json, or any vendored assets/ runtime. See template.json for the upstream metadata snapshot.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 40,144 | 52,118 | +30% | 1 | 1 | 0% | 8,263 | 8,900 | +8% | 0 | 0 | — |
case-02 | fail→fail | 42,563 | 77,995 | +83% | 1 | 1 | 0% | 8,259 | 8,896 | +8% | 0 | 0 | — |
case-03 | fail→pass | 67,579 | 60,351 | -11% | 1 | 1 | 0% | 8,254 | 8,891 | +8% | 0 | 0 | — |
case-04 | pass→fail | 40,077 | 53,831 | +34% | 1 | 1 | 0% | 8,234 | 8,871 | +8% | 0 | 0 | — |
case-05 | pass→fail | 41,130 | 42,772 | +4% | 1 | 1 | 0% | 8,219 | 8,856 | +8% | 0 | 0 | — |
case-06 | pass→fail | 28,420 | 53,468 | +88% | 1 | 1 | 0% | 5,218 | 8,863 | +70% | 0 | 0 | — |
case-07 | fail→fail | 45,293 | 51,908 | +15% | 1 | 1 | 0% | 8,261 | 8,898 | +8% | 0 | 0 | — |
case-08 | fail→fail | 53,346 | 52,421 | -2% | 1 | 1 | 0% | 8,250 | 8,887 | +8% | 0 | 0 | — |
case-09 | fail→fail | 42,345 | 39,694 | -6% | 1 | 1 | 0% | 8,256 | 8,893 | +8% | 0 | 0 | — |
case-10 | fail→fail | 42,076 | 41,515 | -1% | 1 | 1 | 0% | 8,248 | 8,885 | +8% | 0 | 0 | — |
case-11 | fail→fail | 49,616 | 51,964 | +5% | 1 | 1 | 0% | 8,239 | 8,876 | +8% | 0 | 0 | — |
case-12 | fail→pass | 34,718 | 50,355 | +45% | 1 | 1 | 0% | 5,560 | 8,918 | +60% | 0 | 0 | — |
case-13 | fail→fail | 49,127 | 63,068 | +28% | 1 | 1 | 0% | 8,254 | 8,891 | +8% | 0 | 0 | — |
case-14 | fail→pass | 39,041 | 43,461 | +11% | 1 | 1 | 0% | 8,160 | 8,902 | +9% | 0 | 0 | — |
case-15 | fail→pass | 42,928 | 42,191 | -2% | 1 | 1 | 0% | 8,244 | 8,881 | +8% | 0 | 0 | — |
case-16 | fail→fail | 40,973 | 38,288 | -7% | 1 | 1 | 0% | 8,253 | 8,890 | +8% | 0 | 0 | — |
case-17 | fail→pass | 36,825 | 48,965 | +33% | 1 | 1 | 0% | 7,472 | 8,895 | +19% | 0 | 0 | — |
case-18 | fail→fail | 55,068 | 43,146 | -22% | 1 | 1 | 0% | 8,255 | 8,892 | +8% | 0 | 0 | — |
case-19 | fail→fail | 50,719 | 40,249 | -21% | 1 | 1 | 0% | 8,240 | 8,877 | +8% | 0 | 0 | — |
case-20 | fail→pass | 51,158 | 38,642 | -24% | 1 | 1 | 0% | 8,260 | 8,897 | +8% | 0 | 0 | — |
case-21 | fail→fail | 36,492 | 40,382 | +11% | 1 | 1 | 0% | 6,144 | 8,894 | +45% | 0 | 0 | — |
case-22 | fail→fail | 51,897 | 38,410 | -26% | 1 | 1 | 0% | 8,244 | 8,881 | +8% | 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 +14 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
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.