dbs-diagnosis
Automatically determines whether the user is tackling a specific business problem or seeking a comprehensive business model review, routing them into the appropriate diagnostic pipeline. Use when users need to deconstruct a business, audit a business model, or resolve specific business dilemmas.
1How It Works & Mechanism
A pure-prompt business diagnosis skill: its core task is not to answer problems but to 'dissolve' them, grounded in statistics from 8,000+ paid consultations.
**你的核心工作不是回答问题，是消解问题。**
Six core axioms form non-negotiable criteria (objective existence of business models / business models determine ethics / intellect does not monetize directly / traffic does not equal revenue / pricing is the product / 99% are psychological issues).
商业模式是一台有固定 input 要求的机器，人只是喂料员。
Phase 0 automatically identifies the workflow, keeping 'consultation/checkup' as internal tags without prompting the user to choose.
「问诊」和「体检」是内部流程标签，不能要求用户理解或选择。
Consultation mode classifies input by ratio: 10% pure informational (direct answer or navigation), 15% emotional (set boundaries and include in conclusion), and 75% complex problems routed into the dissolution funnel.
### 15% — 情绪宣泄类
The 5-layer dissolution funnel filters progressively and pauses for dialogue at each step, strictly prohibiting all-in-one runs: Linguistic Traps (25%) / Faulty Assumptions (25%) / Logical Fallacies (20%) / Factual Premises (1.5%) / Information Sufficiency (2.5%).
这是 skill 的核心。逐层过滤，每一层都停下来跟用户对话。**不要一次性把所有层跑完。**
Layer 1 sets an actionable criterion: can key terms be quantified or operationally defined? If not, the problem cannot be answered.
**检测方法**：问题中的关键词，能不能给出可量化或可操作的定义？如果不能，这个问题就不可能被回答。
Layer 3 highlights the most common logical fallacy: mistaking correlation for causation (effort vs. results, publishing frequency vs. traffic, survivorship bias).
最常见的错误：把**相关性**当成**因果性**。
The surviving 1% of genuine problems are answered across four categories: Logical Deduction (0.4%), Value Choices (0.3%), Resource Constraints (0.2%), and Out-of-Scope (0.1%, referred to professionals).
活过消解漏斗的 1%，是真正需要被解答的问题。
Checkup mode mandates collecting six business facts upfront (Product / Price / Customer / Acquisition / Delivery / Monthly Revenue), enforcing concreteness via existence checks and the 'Color Test'.
**产品颜色测试**：你能不能说出你的产品是什么颜色的？说不出来就还没进入市场。
The seven checkup audits pause sequentially for user feedback, culminating in a structured diagnostic report (featuring core judgments and a one-sentence prescription).
### 检验 1：印钞机检验
Tracks psychological and thinking-quality signals throughout, surfacing them naturally at most once per signal, without crossing boundaries into psychological intervention.
### 心理问题信号
Mandates a 'Premise Challenge' (inspired by YC office hours) before output, concluding with 'the first thing to do tomorrow' instead of generic advice.
## 前提挑战（借鉴 YC office-hours）
2Core Capabilities
4Recommended Governance & Evidence
- SKILL.md: Pure prompt instructions and rule text; no script execution or external impact.
5Second-Pass Verification
- [ok] 纯 prompt、无脚本 — 目录仅 SKILL.md(21084B) + agents/openai.yaml(269B)；危险 token 扫描零命中。
- [ok] 六条公理与五层漏斗逐字存在 — 公理 1-6 与五层标题（含 25%/25%/20%/1.5%/2.5% 占比）均逐字命中源文。
- [ok] Phase 0 隐藏内部流程 — 原文「「问诊」和「体检」是内部流程标签，不能要求用户理解或选择。」+ 禁止事项三行均在。
- [ok] 七项检验与报告模板 — 检验 1-7 标题齐全；报告模板含「## 基本信息」「## 诊断结果」「## 一句话处方」。
- [ok] 前提挑战与强制行动收尾 — 「## 前提挑战（借鉴 YC office-hours）」整节与强制任务整句逐字命中。
- [ok] 外部依赖为空 — 全文无 URL/CLI/包依赖；YC office-hours 仅作方法来源叙述，未指向外部工具或端点。
- [ok] 无写盘 / 无凭证读取 — 「留档（可选）」一节也只提示用户输入 `/dbs` 由主路由处理，本 skill 不写文件；grep process.env 零命中。
- [ok] description 与内容相符 — description（自动识别具体问题还是全面检查、进入对应诊断流程）对应 Phase 0 + 10%/15%/75% 分类，无夸大；「消解掉 99.1%」的统计口径来自作者自述、无法独立核实，已在 how_it_works 标注为作者口径。
6Verdict & Scenarios
e4fba0e0f79bd8b4…