YS Stage Distiller

Open-source Skill for startup stage diagnosis

先判断阶段,再决定动作。 Stage clarity before GTM.

Stage Distiller 是一个开源 Claude Skill。它帮助 AI builder 在自己的工作环境里,通过多轮深度对话判断:我现在在哪个创业阶段、缺什么证据、什么动作过早、下一步该验证什么。 Stage Distiller is an open-source Claude Skill that helps AI builders diagnose their actual startup stage, missing evidence, premature actions, and next validation step through multi-turn dialogue inside their existing agent workspace.

GTM 是复制一个已经被证明的购买理由,不是用更多渠道掩盖不确定性。 GTM is replication of a proven reason to buy, not noise to mask uncertainty.
Install the Skill Get the framework via email Get the framework via email

About Yuki

一个把个人判断开源成 Skill 的实验。 An experiment in turning founder judgment into a reusable skill.

这里会放一段 Yuki 亲自改写的简短背景:为什么这个 Skill 来自真实创业判断,而不是一套通用创业模板。

This section will carry Yuki’s own short background: why this skill comes from lived startup judgment, not a generic startup template.

The problem

很多 builder 卡住,不是因为做得少,而是把阶段混在一起。 Many builders are stuck because stage boundaries are mixed together.

  • 你说你在做 GTM。你的 MVP 用户还能数得清。You say you are doing GTM. Your MVP users can still be counted by hand.
  • 你在跑 outbound。还没有人重复地为这个问题付过钱。You are running outbound. Nobody has repeatedly paid for the problem yet.
  • 你刚开始招销售。你自己还没用 ICP 拿下过 3 个客户。You are hiring sales. You have not closed three customers through a clear ICP yourself.
  • 你以为你在 PMF Search。其实你在等一个根本不存在的信号。You think you are in PMF Search. You may be waiting for a signal that does not exist.

The six-stage framework

每个阶段都有不同的问题和证据门槛。 Every stage has a different question and evidence gate.

01Idea问题是否真实、具体、值得解决?Is the problem real, specific, and worth solving?
02MVP Definition最小可验证、可交付、可被理解的结果是什么?What is the smallest verifiable outcome?
03MVP Validation有没有人愿意试用、等待、付费或投入时间?Will anyone try, wait, pay, or spend time?
04PMF Search哪个人群、场景和付费理由开始反复成立?Which segment and paid reason repeat?
05GTM Test能不能用一个渠道重复找到相似客户?Can one channel repeatedly find similar customers?
06GTM Scale哪些动作可以被系统化复制和放大?Which motion can be systemized and scaled?
See the full framework on GitHub See the full framework on GitHub

Sample report

输出是一份判断型 Snapshot,不是又一张待办清单。 The output is a diagnostic Snapshot, not another checklist.

Google Ads misfire test

Current Stage: MVP Validation (confidence: medium-high)

Evidence Gap:
- non-friend users
- paid or prepaid willingness
- clear ICP and repeated use case

Misfire Risk:
Premature Google Ads spend before user/payment evidence.

Next validation:
Find 5-10 independent users and test one narrow MVP promise.
Read full samples Read full samples

How it works

在你工作的 agent 环境里运行,而不是再开一个网页工具。 Run it inside your agent workspace, not another web app.

Install

把 Skill 安装到 Claude Code、Cursor 或 Codex 的本地 skills 目录。Install the skill into Claude Code, Cursor, or Codex.

Invoke

告诉 agent:帮我判断当前创业阶段、PMF、GTM readiness 或验证门槛。Ask for stage diagnosis, PMF, GTM readiness, or validation gates.

Dialogue

Skill 会追问具体用户、已发生行为、证据强度和过早动作。The skill probes concrete users, past behavior, evidence strength, and premature actions.

Snapshot

输出并保存 Stage Snapshot,再选择是否继续完整报告、阶段深挖或 7 天验证动作。It saves a Stage Snapshot, then offers a full report, deep dive, or 7-day validation actions.

Get the framework

留下邮箱,等第一版 framework / workbook / sample report 发出。 Leave your email for the first framework, workbook, and sample report.

Why this exists

把只有少数人能获得的判断,变成更多 builder 能调用的容器。 Turning scarce judgment into a container more builders can run.

这里会放 Yuki 亲自写的故事:为什么不是再做一个 SaaS,而是先把阶段判断开源成一个 Skill。

This section will carry Yuki’s story: why this starts as an open-source skill rather than another SaaS dashboard.