Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →生成 Ian 风格的中文正文配图。用于用户要求为中文文章、帖子、博客、Notion 文档、工作流文档、方法论、流程、结构、状态、隐喻或观点生成“怪诞”“小黑”“手绘”“正文配图”“文章插图”“配图建议”“shot list”“去标题/改图”等任务；默认使用小黑 IP、纯白手绘、少量红橙蓝批注、简洁清爽但天马行空的视觉风格。
.claude/skills/chaitanyagiri-ian-xiaohei-illustrations/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -27% | 0% |
为中文文章设计和生成 16:9 横版正文配图。目标不是做商业插画、PPT 信息图或可爱卡通，而是把文章里的关键判断、流程、结构、状态或隐喻，变成一张清爽、怪诞、有创意、可读但不说明书的手绘解释图。
默认视觉 IP 是“小黑”：黑色实心、白点眼、细腿、空表情，认真做一件荒诞但成立的事。小黑必须参与画面的核心动作，不能只是站在旁边当装饰。
按任务需要读取，不要一次塞满上下文：
references/style-dna.md：风格 DNA、颜色、文字、禁忌。references/xiaohei-ip.md：小黑 IP 的形象、性格、动作库和禁忌。references/composition-patterns.md：结构类型、原创隐喻方法和反复刻规则。references/prompt-template.md：单张生图提示词模板。references/qa-checklist.md：生成后检查和迭代规则。assets/examples/：只作低频视觉校准，不进入默认生成路径。不要照抄这些案例的构图、物件或标注。先读用户给的正文、链接、Notion 页面、Markdown 文件或截图内容。提炼：
不要平均配图。优先选择“认知锚点”，例如：核心判断、两个断点、输入输出闭环、分流、前后对比、一鱼多吃、承接路径、常见坑、角色状态变化。
如果用户只是说“分析怎么配图 / 思考哪些地方需要配图”，先给 shot list。每张图写清楚：
默认 4-8 张。文章很短时 1-3 张；长文也不要轻易超过 9 张。够用就好，避免把正文做成画册。
如果用户明确要求“生成 / 输出 / 做图 / 帮我生成”，不要停下来等确认；用内置 image_gen 每张单独生成。不要把多张图拼在一张里。
每张图只讲一个核心结构。提示词必须包含：
不要复刻过往案例。案例只提供风格密度和小黑参与方式，不能直接复用“传送带断点 / 小黑拉线 / 素材鱼 / 盖章工具箱 / 常见坑路径”等已有构图，除非用户明确要求复刻某张图。每次都要从当前文章重新发明一个奇怪但成立的隐喻。
生成后检查 references/qa-checklist.md。如果出现以下问题，优先重生成或局部编辑：
如果用户在 workspace 内工作，把最终图复制到：
textassets/<article-slug>-illustrations/
按顺序命名：
text01-topic-name.png 02-topic-name.png
保留原始生成文件，不要覆盖已有资产，除非用户明确要求替换。
生成前的策略输出要短而准。生成后的交付要包含：
不要长篇解释风格理论；让图自己说话。
These project-level overrides apply on top of everything above whenever this skill runs in this repo:
blog/src/assets/media/<post-slug>/ and each generated file must be recorded in blog/src/_data/media.json (set the entry's status to "ready").Attribution: this skill is "Ian Xiaohei Illustrations" by Ian (https://github.com/helloianneo), MIT — see LICENSE and NOTICE.md in this folder.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 20,516 | 22,526 | +10% | 1 | 1 | 0% | 3,313 | 4,618 | +39% | 0 | 0 | — |
case-02 | fail→fail | 39,894 | 21,081 | -47% | 1 | 1 | 0% | 7,892 | 1,507 | -81% | 0 | 0 | — |
case-03 | fail→fail | 5,037 | 6,573 | +30% | 1 | 1 | 0% | 286 | 1,610 | +463% | 0 | 0 | — |
case-04 | fail→pass | 15,998 | 6,533 | -59% | 1 | 1 | 0% | 2,223 | 2,226 | +0% | 0 | 0 | — |
case-05 | pass→pass | 32,876 | 4,241 | -87% | 1 | 1 | 0% | 1,877 | 1,748 | -7% | 0 | 0 | — |
case-06 | fail→pass | 19,470 | 5,175 | -73% | 1 | 1 | 0% | 2,194 | 1,985 | -10% | 0 | 0 | — |
case-07 | fail→pass | 11,071 | 3,416 | -69% | 1 | 1 | 0% | 1,818 | 1,734 | -5% | 0 | 0 | — |
case-08 | fail→pass | 15,143 | 3,581 | -76% | 1 | 1 | 0% | 2,326 | 1,688 | -27% | 0 | 0 | — |
case-09 | pass→pass | 18,051 | 23,304 | +29% | 1 | 1 | 0% | 1,673 | 3,304 | +97% | 0 | 0 | — |
case-10 | pass→pass | 13,334 | 3,252 | -76% | 1 | 1 | 0% | 1,771 | 1,678 | -5% | 0 | 0 | — |
case-11 | fail→fail | 11,299 | 9,616 | -15% | 1 | 1 | 0% | 1,599 | 2,502 | +56% | 0 | 0 | — |
case-12 | pass→pass | 10,975 | 10,660 | -3% | 1 | 1 | 0% | 1,648 | 2,345 | +42% | 0 | 0 | — |
case-13 | fail→pass | 13,007 | 34,291 | +164% | 1 | 1 | 0% | 1,627 | 2,680 | +65% | 0 | 0 | — |
case-14 | pass→pass | 40,183 | 5,376 | -87% | 1 | 1 | 0% | 1,654 | 2,039 | +23% | 0 | 0 | — |
case-15 | pass→pass | 8,569 | 7,661 | -11% | 1 | 1 | 0% | 1,245 | 2,164 | +74% | 0 | 0 | — |
case-16 | fail→pass | 28,372 | 26,143 | -8% | 1 | 1 | 0% | 1,303 | 5,812 | +346% | 0 | 0 | — |
case-17 | pass→pass | 15,664 | 9,693 | -38% | 1 | 1 | 0% | 1,889 | 2,599 | +38% | 0 | 0 | — |
case-18 | fail→pass | 13,658 | 5,002 | -63% | 1 | 1 | 0% | 1,855 | 2,122 | +14% | 0 | 0 | — |
case-19 | pass→fail | 21,138 | 31,350 | +48% | 1 | 1 | 0% | 3,294 | 6,370 | +93% | 0 | 0 | — |
case-20 | pass→pass | 64,918 | 26,321 | -59% | 1 | 1 | 0% | 8,229 | 6,311 | -23% | 0 | 0 | — |
case-21 | pass→pass | 17,861 | 29,046 | +63% | 1 | 1 | 0% | 2,778 | 3,818 | +37% | 0 | 0 | — |
case-22 | fail→pass | 25,692 | 7,878 | -69% | 1 | 1 | 0% | 1,600 | 2,341 | +46% | 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 20 counted toward the lift figure. The other 2 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 +36 percentage points is the difference between those two pass rates over the 20 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.