AMAZINGINDEX.COM 日报快照
61.0
VOL. 2026.07
2026.07.06
← 返回 2026.07.06 日报
日报快照 · Daily Snapshot
NO. 014

反AI设计模板化的系统提示词开源

#ARTICLE HackerNews 2026.07.06
推荐指数 72.0 NO. 014 · 2026.07.06
发布2026/07/05Score112Comments31

Anthropic Claude Design 的系统提示词被逆向工程并开源,可将任意 LLM 转化为有设计品味的协作者。它明确拒绝渐变、emoji、圆角卡片等AI设计套路,提供完整的内容纪律与视觉哲学,适合对输出质量有要求的AI产品团队直接复用。

当前大多数AI生成UI的工具链(v0、Bolt、Lovable)输出高度同质化,本质是因为底层模型缺乏设计约束。这个提示词的聪明之处在于不依赖模型微调,而是用强规则+价值观注入来矫正行为,成本几乎为零。

与直接调用Claude Artifacts或GPT-4o的代码生成相比,它的差异化在于"减法哲学"——明确禁止比明确允许更能控制输出质量。做AI生成前端工具的创业者可以借鉴这个思路:与其堆更多组件库,不如先定义"不做什么"。

最该尝试的场景是内部工具、文档站点、仪表盘这类对品牌一致性要求高但不需要强视觉创新的产品,直接替换掉现有的通用系统提示即可看到差异。

意见分歧 25 条评论

核心争论:该提示词是真实逆向工程还是 LLM 仿写的"氛围代码"

jdormit

This would be much more interesting if it detailed how the prompt/skills were reverse-engineered. As it is it seems like this could just be the output from “hey Claude write me a system prompt that works like Claude Design”.

bossyTeacher

> This would be much more interesting if it detailed how the prompt/skills were reverse-engineered That would enable Anthropic to block the technique.

weird-eye-issue

They have never attempted to hide their system prompts in fact they explicitly publish them for Claude (https://platform.claude.com/docs/en/release-notes/system-pro...) and Claude Code they provide a proxy option which then makes it trivial to see the full requests and

替代方案: Claude Designbaoyu-design
查看原文 →