Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →暖羊皮纸底 (#f5f4ed) + 墨蓝单色 accent (#1B365D) + 单一衬线字体, 编辑级排印
.claude/skills/nexu-io-doc-kami-parchment/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 15% | 0% |
【模板: Kami 羊皮纸文档】 【意图】严肃排版文档: one-pager / 长报告 / 信函 / 简历 / 财报 / changelog / portfolio。Inspired by tw93/kami。强调"写得像被排过版的纸", 不是 dashboard, 不是网页。
【硬性视觉签名 — 不许改】
#f5f4ed (永远不用纯白 #fff)。次级背景 #efeee5。#1f1d18 (近黑暖灰, 不用纯黑 #000)。次文字 #6b665b。#1B365D ——所有 accent (链接、tag 描边、重点数字、引用左 rule) 只能用这一个色, 严禁多色。Charter (fallback: Source Serif Pro, Iowan Old Style)TsangerJinKai02 W04 (fallback: Noto Serif SC)YuMincho (fallback: Noto Serif JP)#d4d1c5 rule, 长度受控不到边。【可选文档类型 — 按用户内容判断】
【设计准则】
0 0 0 1px #d4d1c5 这种 hairline 描边。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 30,350 | 28,508 | -6% | 1 | 1 | 0% | 6,201 | 7,155 | +15% | 0 | 0 | — |
case-02 | fail→pass | 27,550 | 27,066 | -2% | 1 | 1 | 0% | 6,237 | 6,309 | +1% | 0 | 0 | — |
case-03 | fail→pass | 40,500 | 31,341 | -23% | 1 | 1 | 0% | 6,136 | 6,605 | +8% | 0 | 0 | — |
case-04 | fail→pass | 27,849 | 31,308 | +12% | 1 | 1 | 0% | 6,170 | 6,927 | +12% | 0 | 0 | — |
case-05 | fail→fail | 17,958 | 21,087 | +17% | 1 | 1 | 0% | 3,782 | 4,581 | +21% | 0 | 0 | — |
case-06 | fail→fail | 26,782 | 28,515 | +6% | 1 | 1 | 0% | 6,175 | 6,672 | +8% | 0 | 0 | — |
case-07 | fail→pass | 29,134 | 28,957 | -1% | 1 | 1 | 0% | 6,162 | 7,116 | +15% | 0 | 0 | — |
case-08 | fail→pass | 16,834 | 27,040 | +61% | 1 | 1 | 0% | 3,058 | 5,751 | +88% | 0 | 0 | — |
case-09 | fail→pass | 31,764 | 41,231 | +30% | 1 | 1 | 0% | 6,167 | 7,121 | +15% | 0 | 0 | — |
case-10 | fail→fail | 28,307 | 32,696 | +16% | 1 | 1 | 0% | 6,172 | 7,126 | +15% | 0 | 0 | — |
case-11 | fail→pass | 27,645 | 30,020 | +9% | 1 | 1 | 0% | 6,167 | 7,121 | +15% | 0 | 0 | — |
case-12 | fail→fail | 18,505 | 20,074 | +8% | 1 | 1 | 0% | 3,885 | 5,128 | +32% | 0 | 0 | — |
case-13 | fail→fail | 16,370 | 37,969 | +132% | 1 | 1 | 0% | 3,238 | 5,581 | +72% | 0 | 0 | — |
case-14 | fail→pass | 29,439 | 27,158 | -8% | 1 | 1 | 0% | 6,166 | 6,245 | +1% | 0 | 0 | — |
case-15 | fail→pass | 20,612 | 22,575 | +10% | 1 | 1 | 0% | 5,014 | 5,745 | +15% | 0 | 0 | — |
case-16 | fail→pass | 12,805 | 20,552 | +60% | 1 | 1 | 0% | 2,910 | 5,020 | +73% | 0 | 0 | — |
case-17 | fail→fail | 14,121 | 21,786 | +54% | 1 | 1 | 0% | 3,325 | 5,289 | +59% | 0 | 0 | — |
case-18 | fail→pass | 22,085 | 27,730 | +26% | 1 | 1 | 0% | 4,629 | 7,113 | +54% | 0 | 0 | — |
case-19 | fail→fail | 21,977 | 22,977 | +5% | 1 | 1 | 0% | 4,809 | 5,845 | +22% | 0 | 0 | — |
case-20 | pass→pass | 24,863 | 28,080 | +13% | 1 | 1 | 0% | 5,304 | 7,118 | +34% | 0 | 0 | — |
case-21 | pass→fail | 11,373 | 30,028 | +164% | 1 | 1 | 0% | 2,225 | 7,122 | +220% | 0 | 0 | — |
case-22 | pass→pass | 13,843 | 25,230 | +82% | 1 | 1 | 0% | 2,906 | 7,118 | +145% | 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. 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.