Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build research slides with text-first source (Slidev/Marp/Reveal/Quarto) and reproducible export (PDF). Includes structure, figure reuse rules, and quality checklist for top-tier scientific presentations.
.claude/skills/foryourhealth111-pixel-slides-as-code/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 52 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -42% | 0% |
解决常见问题：
本 skill 让 slides 拥有：
figures/ 最终导出，不截图）1) Slidev：开发者友好，组件化，适合组会/技术分享 2) Marp：排版可控，适合快速生成简洁风格 3) Reveal.js：网页交互强，适合 demo/交互展示 4) Quarto slides：报告/论文/幻灯片一体化（推荐用于报告+slides联动）
> 内容结构建议使用 scientific-slides；本 skill 更偏“工程化与导出管线”。
figures/**/out/ 引用（PDF/SVG/PNG）推荐骨架： 1) 背景与缺口（1–2 页） 2) 方法直觉（1–2 页示意图） 3) 核心结果（3–6 页，每页一图一结论） 4) 局限与下一步（1–2 页）
参考：templates/slidev-structure.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 | 17,342 | 16,759 | -3% | 1 | 1 | 0% | 2,895 | 3,281 | +13% | 0 | 0 | — |
case-02 | fail→pass | 22,628 | 20,970 | -7% | 1 | 1 | 0% | 3,963 | 4,297 | +8% | 0 | 0 | — |
case-03 | fail→pass | 15,567 | 15,921 | +2% | 1 | 1 | 0% | 2,820 | 3,028 | +7% | 0 | 0 | — |
case-04 | fail→pass | 14,426 | 11,556 | -20% | 1 | 1 | 0% | 2,403 | 2,404 | +0% | 0 | 0 | — |
case-05 | pass→pass | 10,497 | 9,661 | -8% | 1 | 1 | 0% | 1,622 | 1,940 | +20% | 0 | 0 | — |
case-06 | pass→pass | 11,458 | 11,164 | -3% | 1 | 1 | 0% | 1,961 | 2,410 | +23% | 0 | 0 | — |
case-07 | pass→fail | 12,638 | 15,340 | +21% | 1 | 1 | 0% | 2,298 | 3,241 | +41% | 0 | 0 | — |
case-08 | pass→pass | 12,739 | 13,420 | +5% | 1 | 1 | 0% | 2,199 | 2,674 | +22% | 0 | 0 | — |
case-09 | fail→pass | 10,665 | 4,308 | -60% | 1 | 1 | 0% | 1,842 | 1,179 | -36% | 0 | 0 | — |
case-10 | pass→pass | 10,981 | 7,248 | -34% | 1 | 1 | 0% | 1,840 | 1,601 | -13% | 0 | 0 | — |
case-11 | pass→pass | 14,292 | 11,945 | -16% | 1 | 1 | 0% | 2,355 | 2,647 | +12% | 0 | 0 | — |
case-12 | pass→pass | 10,546 | 8,755 | -17% | 1 | 1 | 0% | 1,659 | 1,910 | +15% | 0 | 0 | — |
case-13 | pass→pass | 12,950 | 9,995 | -23% | 1 | 1 | 0% | 2,132 | 1,986 | -7% | 0 | 0 | — |
case-14 | pass→pass | 15,879 | 17,797 | +12% | 1 | 1 | 0% | 2,738 | 3,307 | +21% | 0 | 0 | — |
case-15 | pass→pass | 14,675 | 2,196 | -85% | 1 | 1 | 0% | 2,460 | 729 | -70% | 0 | 0 | — |
case-16 | pass→pass | 12,579 | 10,655 | -15% | 1 | 1 | 0% | 2,054 | 2,185 | +6% | 0 | 0 | — |
case-17 | fail→pass | 9,476 | 2,714 | -71% | 1 | 1 | 0% | 1,395 | 803 | -42% | 0 | 0 | — |
case-18 | fail→pass | 8,537 | 5,574 | -35% | 1 | 1 | 0% | 1,488 | 1,287 | -14% | 0 | 0 | — |
case-19 | pass→pass | 9,326 | 2,868 | -69% | 1 | 1 | 0% | 1,382 | 825 | -40% | 0 | 0 | — |
case-20 | fail→pass | 7,731 | 1,622 | -79% | 1 | 1 | 0% | 1,277 | 605 | -53% | 0 | 0 | — |
case-21 | pass→pass | 6,650 | 4,632 | -30% | 1 | 1 | 0% | 1,191 | 1,143 | -4% | 0 | 0 | — |
case-22 | pass→pass | 6,967 | 4,545 | -35% | 1 | 1 | 0% | 1,217 | 1,290 | +6% | 0 | 0 | — |
case-23 | pass→pass | 1,960 | 1,915 | -2% | 1 | 1 | 0% | 316 | 712 | +125% | 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. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is 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.