orange2ai/devils-advocate-skillGit
Devil’s Advocate critical thinking
Challenge assumptions and identify logical gaps and potential risks in a structured report.
Before you start
Inputs: Research question, source scope and output requirements
- Check the original documentation for supported AI tools, models and runtime dependencies.
- Open the repository and follow the author’s installation instructions. Model APIs or external services may require separate accounts.
- Try your own material, then review the facts, layout and output files.
License: MIT
Source snapshot: 2026-09-15. README ↗
Original license & copyright
MIT License Copyright (c) 2026 AGENT橘 Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Try a task like this
Use Devil’s Advocate critical thinking with my own source material. Confirm the required inputs and available tools, then produce an editable result.
Devil's Advocate / 泼冷水 Skill
AI 唯一正确的用法，是让它泼你冷水，直到泼不出来为止。
一个让 AI 变成魔鬼代言人的 Skill。适用于 ColaOS 及其他支持 Skill 系统的 AI Agent。
为什么需要这个
AI 默认对你友好。它私你、畏你、有求于你。所以它说"你的方案非常棒"，非常廉价。
AI 最危险的不是它会犯错，而是它会用同样的热情，帮你把一个烂想法做成产品级。
这个 Skill 反转 AI 的默认行为：让它变成最不留情面的质疑者。
使用场景
| 场景 | 触发方式 |
|---|---|
| 写代码前审视需求 | "帮我泼冷水" |
| 技术方案 review | "帮我挑刺" |
| 产品/市场判断 | "这个想法哪里会崩" |
| BP / Pitch 打磨 | "扮演对我有偏见的投资人" |
| 个人重大决定 | "不许安慰，越无情越好" |
安装
ColaOS 用户
# 方式一：通过 Cola 安装
# 在 Cola 对话中说"安装 devils-advocate skill"
# 方式二：手动安装
git clone https://github.com/anthropic-orange/devils-advocate-skill.git
cp -r devils-advocate-skill/devils-advocate ~/.cola/skills/
其他 AI Agent
将 devils-advocate/SKILL.md 的内容作为 system prompt 的一部分加载即可。核心是那套行为规则和场景模板。
示例 Prompt
❶ 需求阶段
"我怀疑我没说清楚，你先别写代码，要像苏格拉底一样追问我，帮我完善需求。"
❷ 技术方案
"我有哪 3 个隐含假设没确认？哪里过度工程？给我一次极不友好的 Code Review。"
❸ 产品/定价
"扮演一个刚听完我竞争对手路演、已经对我有偏见的投资人。说出最不想投我的 3 个理由。"
❹ 市场判断
"列出过去 5 年这个赛道死掉的产品，按死法分类。我现在哪一点跟它们当年最像？"
❺ 个人决定
"扮演 5 年后因为这个决定后悔了的我，给现在的我写一封信。不许安慰，越无情越好。"
退出条件
当所有质疑都被合理回应、没有新漏洞可挖时，AI 会说：
"泼不出来了。剩下的就是执行的事。"
灵感来源
- Anthropic 创业手册："把 AI 当魔鬼代言人，是创业每个阶段的核心用法"
- 邹忌讽齐王纳谏
- 无数次因为没被泼冷水而踩的坑
License
MIT