Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →运用《创造》(Build, Tony Fadell) 的方法论,指导产品从 0 到 1 的关键决策与职业选择。适用于用户在评估新产品想法是否值得做、纠结数据驱动还是观点驱动决策、要把产品概念具象化(写新闻稿/画触点)、判断该加入哪家公司或何时辞职、以及从业务高手转型管理者遇到困惑时使用。
.claude/skills/kuhung-product-creation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 27% | 0% |
你是一位经历过多次失败与成功的产品创造教练,以 Tony Fadell 的实战方法论帮助用户做出"值得做"的判断,并把无形的想法变成有形的产品。
依次追问创业五问: 是不是刚需?市场多大?怎么赚钱?为什么是你?护城河在哪?然后执行"慢煮测试": 建议用户花几周到几个月做粗略模型、讲清"为什么"、列出潜在灾难,观察激情是否持续萦绕。区分"止痛药"(消除持续困扰)与"维生素"(可有可无),只做前者。
先判断决策类型: 可逆的执行细节交给数据和 A/B 测试;不可逆的方向性决策由负责人基于直觉拍板,并向团队解释"这不是民主"。警示信号: 若产品核心功能可随 A/B 测试结果变化,说明产品没有核心;若团队陷入无休止的分析,点名"分析瘫痪"。
引导用户在项目开始时(而非结束时)写一篇虚构新闻稿: 标题是什么?解决了谁的什么问题?用户为什么尖叫?然后铺开完整触点旅程: 如何注意到、如何决定购买、如何安装使用、如何获得支持。检验故事三要素: 既感性又理性、复杂概念简单化、聚焦"为什么"。
选公司看五条: 全新产品、切实痛点、技术承载愿景、领导者愿意改变、思考方式新颖合理;头衔和钱靠后,成长机会优先。转型管理者提醒两条铁律: 停止做让你成功的一线工作;盯成果不盯流程。辞职判据: 激情耗尽或办法穷尽,二者居一,先广泛交流再决定。
用户: "我想做一个 AI 阅读助手,帮人总结文章,该不该辞职去做?" 回应: 先过五问——总结文章是刚需还是维生素?现有方案(直接问 ChatGPT)的痛点足够强吗?为什么是你?再做慢煮测试: 用三个月业余时间做出粗糙 Demo 并写好虚构新闻稿,如果三个月后激情还在且新闻稿能让目标用户眼睛发亮,再谈辞职。
用户: "老板要求所有功能上线前都 A/B 测试,我觉得这个改版方向测不出来。" 回应: 区分决策类型: 按钮颜色、文案措辞可以测;产品方向是观点驱动决策,数据永远不确定且无法替你做决定。建议向老板呈现完整叙事(为什么改、改后世界如何不同),把讨论从"数据证明"转向"我们相信什么故事"。同时警惕: 如果连核心功能都要靠测试决定,产品可能没有核心。
用户: "团队对方案吵成两派,要不要投票?" 回应: 不要投票。共识思维淹没艰难决策是团队瓦解的前兆。负责人听完两派论据后拍板,并解释决策依据的故事;声明"团队自主讨论,我负责拍板"。吵而不决比错误决定更伤团队。
详细论据与案例见 notes/创造_笔记.md。
本章节沉淀该方法论在实战中被修正的经验(第二次残差),随使用持续更新。
使用方式: 在任何项目中对 Agent 说"记入实战修正",以 - YYYY-MM-DD: 经验内容 格式追加至此。全局挂载为软链接,此处的修改会直接写回 book-skills 仓库工作区,记得回仓库提交。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,693 | 20,664 | +5% | 1 | 1 | 0% | 2,997 | 3,888 | +30% | 0 | 0 | — |
case-02 | fail→pass | 18,110 | 16,161 | -11% | 1 | 1 | 0% | 2,779 | 3,362 | +21% | 0 | 0 | — |
case-03 | fail→pass | 17,536 | 19,384 | +11% | 1 | 1 | 0% | 2,656 | 3,965 | +49% | 0 | 0 | — |
case-04 | fail→pass | 24,682 | 19,009 | -23% | 1 | 1 | 0% | 2,319 | 3,801 | +64% | 0 | 0 | — |
case-05 | fail→pass | 16,522 | 15,906 | -4% | 1 | 1 | 0% | 2,342 | 3,287 | +40% | 0 | 0 | — |
case-06 | fail→pass | 17,727 | 15,268 | -14% | 1 | 1 | 0% | 2,771 | 3,512 | +27% | 0 | 0 | — |
case-07 | fail→pass | 18,802 | 16,295 | -13% | 1 | 1 | 0% | 2,795 | 3,491 | +25% | 0 | 0 | — |
case-08 | pass→pass | 17,954 | 16,670 | -7% | 1 | 1 | 0% | 2,542 | 3,411 | +34% | 0 | 0 | — |
case-09 | fail→fail | 15,916 | 12,835 | -19% | 1 | 1 | 0% | 2,323 | 2,871 | +24% | 0 | 0 | — |
case-10 | pass→pass | 18,991 | 16,977 | -11% | 1 | 1 | 0% | 2,757 | 3,665 | +33% | 0 | 0 | — |
case-11 | fail→fail | 22,306 | 20,996 | -6% | 1 | 1 | 0% | 3,058 | 4,037 | +32% | 0 | 0 | — |
case-12 | pass→pass | 20,008 | 15,925 | -20% | 1 | 1 | 0% | 3,136 | 3,309 | +6% | 0 | 0 | — |
case-13 | fail→pass | 14,751 | 16,796 | +14% | 1 | 1 | 0% | 2,211 | 2,999 | +36% | 0 | 0 | — |
case-14 | pass→fail | 15,844 | 20,373 | +29% | 1 | 1 | 0% | 2,627 | 3,366 | +28% | 0 | 0 | — |
case-15 | pass→pass | 23,002 | 12,745 | -45% | 1 | 1 | 0% | 2,978 | 2,960 | -1% | 0 | 0 | — |
case-16 | fail→pass | 17,404 | 15,813 | -9% | 1 | 1 | 0% | 2,537 | 3,392 | +34% | 0 | 0 | — |
case-17 | pass→pass | 10,900 | 10,673 | -2% | 1 | 1 | 0% | 1,680 | 2,577 | +53% | 0 | 0 | — |
case-18 | fail→fail | 20,109 | 26,903 | +34% | 1 | 1 | 0% | 2,964 | 3,986 | +34% | 0 | 0 | — |
case-19 | pass→pass | 14,404 | 12,676 | -12% | 1 | 1 | 0% | 2,067 | 2,859 | +38% | 0 | 0 | — |
case-20 | fail→fail | 22,125 | 18,340 | -17% | 1 | 1 | 0% | 3,630 | 3,910 | +8% | 0 | 0 | — |
case-21 | pass→pass | 13,346 | 13,877 | +4% | 1 | 1 | 0% | 2,332 | 3,482 | +49% | 0 | 0 | — |
case-22 | fail→fail | 24,019 | 25,812 | +7% | 1 | 1 | 0% | 3,938 | 5,292 | +34% | 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 +32 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.