Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →分析文章内容，在需要视觉辅助理解的位置生成插画。配图可以是信息补充、概念具象化，或引导读者想象。当用户要求"给文章配图"、"为文章生成插图"、"添加配图"时使用此技能。
.claude/skills/anbeime-article-illustrator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 124% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1569% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 35% | 0% |
分析文章结构与内容，识别需要视觉辅助的位置，生成风格灵活的配图。
用户明确要求：
根据用户提供的信息获取文章：
逐段分析文章，识别需要配图的位置。
配图的三种作用：
适合配图的内容：
不需要配图的内容：
配图数量（按文章长度）：
为每个配图位置创建结构化计划：
markdown**配图 1** **插入位置**：[章节名称] / [段落描述] **配图目的**：[为什么这里需要配图] **视觉内容**：[图片应该展示什么] **文件名**：illustration-[slug].png
文件命名规则：
illustration-[slug].pngillustration-product-evolution.png、illustration-ai-vs-human.png根据配图计划，为每张图片生成详细的视觉描述并创建图片。
生成要求：
references/style-guide.md）图片属性：
将生成的图片插入到文章对应位置。
插入规则：
图片保存位置：
imgs/ 子目录完成所有配图后，输出汇总信息：
配图完成！
文章：[文章路径]
生成数量：X/N 张成功
配图位置：
- illustration-[slug].png → [章节/段落位置]
[如有失败]
失败项：
- illustration-[slug].png：[失败原因]识别关键位置：
优先级排序：
风格选择原则：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 18,670 | 17,756 | -5% | 1 | 1 | 0% | 1,975 | 4,416 | +124% | 0 | 0 | — |
case-02 | fail→fail | 16,764 | 12,963 | -23% | 1 | 1 | 0% | 2,946 | 2,921 | -1% | 0 | 0 | — |
case-03 | fail→pass | 6,392 | 29,667 | +364% | 1 | 1 | 0% | 393 | 6,558 | +1569% | 0 | 0 | — |
case-04 | fail→pass | 10,171 | 7,870 | -23% | 1 | 1 | 0% | 1,707 | 2,274 | +33% | 0 | 0 | — |
case-05 | pass→pass | 11,807 | 3,511 | -70% | 1 | 1 | 0% | 1,915 | 1,818 | -5% | 0 | 0 | — |
case-06 | fail→pass | 8,183 | 4,241 | -48% | 1 | 1 | 0% | 1,489 | 1,773 | +19% | 0 | 0 | — |
case-07 | fail→pass | 15,564 | 6,865 | -56% | 1 | 1 | 0% | 1,893 | 2,561 | +35% | 0 | 0 | — |
case-08 | pass→pass | 20,333 | 18,854 | -7% | 1 | 1 | 0% | 2,444 | 3,532 | +45% | 0 | 0 | — |
case-09 | fail→pass | 17,096 | 18,718 | +9% | 1 | 1 | 0% | 2,570 | 3,399 | +32% | 0 | 0 | — |
case-10 | pass→pass | 14,738 | 7,193 | -51% | 1 | 1 | 0% | 2,460 | 2,539 | +3% | 0 | 0 | — |
case-11 | fail→pass | 13,288 | 7,996 | -40% | 1 | 1 | 0% | 2,081 | 2,471 | +19% | 0 | 0 | — |
case-12 | fail→pass | 14,952 | 9,178 | -39% | 1 | 1 | 0% | 2,016 | 2,592 | +29% | 0 | 0 | — |
case-13 | fail→pass | 12,758 | 7,946 | -38% | 1 | 1 | 0% | 2,317 | 2,665 | +15% | 0 | 0 | — |
case-14 | pass→pass | 6,668 | 4,307 | -35% | 1 | 1 | 0% | 1,306 | 2,034 | +56% | 0 | 0 | — |
case-15 | pass→pass | 11,685 | 8,745 | -25% | 1 | 1 | 0% | 1,950 | 2,575 | +32% | 0 | 0 | — |
case-16 | pass→pass | 11,436 | 10,482 | -8% | 1 | 1 | 0% | 1,853 | 2,548 | +38% | 0 | 0 | — |
case-17 | pass→pass | 16,964 | 11,543 | -32% | 1 | 1 | 0% | 2,745 | 3,147 | +15% | 0 | 0 | — |
case-18 | pass→pass | 14,786 | 10,544 | -29% | 1 | 1 | 0% | 2,023 | 2,895 | +43% | 0 | 0 | — |
case-19 | fail→fail | 7,232 | 4,274 | -41% | 1 | 1 | 0% | 992 | 1,902 | +92% | 0 | 0 | — |
case-20 | fail→fail | 5,145 | 1,897 | -63% | 1 | 1 | 0% | 865 | 1,448 | +67% | 0 | 0 | — |
case-21 | fail→fail | 10,036 | 7,946 | -21% | 1 | 1 | 0% | 1,989 | 2,681 | +35% | 0 | 0 | — |
case-22 | pass→pass | 9,264 | 2,538 | -73% | 1 | 1 | 0% | 1,585 | 1,651 | +4% | 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 21 counted toward the lift figure. The other 1 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 +41 percentage points is the difference between those two pass rates over the 21 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.