Name: Cognition Shaper
Author: Victor42

能力

对话长什么样

你说：「看到新闻说芯片出口又被限制，显卡跟着涨价，想搞懂这条链」
侧边框架原位长出骨架——
物理底座：先进制程要用极紫外光刻，光源、镜片、光刻胶的供应商就那么几家
产业层：设计 → 代工 → 封装 → 渠道，每一段的产能与库存周期都不一样
制度层：出口管制清单按算力密度划线，不按产品名
待验证：这轮涨价是产能不够，还是渠道惜售？看新旧卡是否同涨、交期是否同步拉长
它只回一两句：「新卡和旧卡是一起涨，还是只有新卡涨？只有新卡涨指向产能与清单，全都涨更像渠道在囤。」
侧边那份文档你看得见：原生侧边栏、你指定的本地 Markdown，或云端一个未公开列出的 Gist

什么时候值得开

要弄懂一个陌生领域或一件新事物，而不是查一条事实
一件事反复想不透，需要被一层层问下去
要跟 AI 多轮共建一个理解，而不是问一句拿一个答案
想把过程中的假设、反例、已确认细节留下来
只有你明确要求时才启动，普通提问不会变成大工程

它跟普通聊天差在哪

聊天保持短促：每轮最多推进一个因果问题，不甩一大段综述
侧边文档原位展开：不写流水账，已确认的细节与未定的猜测分开放，不丢
从具体现象切入，向下追到第一性原理，并标出跨层的因果

边界与隐私

不用于一次性解释、翻译或内容发布
云端走 secret Gist：未公开列出，但并不等于加密；涉及敏感内容它会先问你
本地模式只写你指定的那个文件，不自己挑路径

Capabilities

What a session looks like

You say: “A headline says chip exports are restricted again and GPU prices are climbing — I want to understand the chain.”
The side surface grows its skeleton in place —
Physics: leading-edge nodes need EUV lithography, and the light sources, mirrors and photoresists come from a handful of suppliers
Industry: design → foundry → packaging → channel, each stage on its own capacity and inventory cycle
Rules: export lists draw the line by compute density, not by product name
Open question: is this rise capacity or channel hoarding? Compare whether old and new cards move together, and whether lead times stretch in step
It replies in one or two sentences: “Are old and new cards rising together, or only the new ones? Only the new ones point at capacity and the lists; everything rising looks more like hoarding.”
You can watch that document: a native side panel, a local Markdown file you name, or one unlisted Gist in the cloud

When it earns its place

You need to understand an unfamiliar field or a new thing, not look up one fact
A question keeps resisting you and needs to be asked down, layer by layer
You want to co-build an understanding over many turns, not collect one-shot answers
You want the hypotheses, counterexamples and confirmed details to survive the conversation
It only starts when you ask for it; an ordinary question won't turn into this

How it differs from ordinary chat

Chat stays short: at most one causal question per turn, no wall of summary
The side document is rewritten in place: never a changelog, and confirmed detail or open uncertainty is never dropped
It enters through a concrete phenomenon, digs down to first principles, and marks causal edges that cross layers

Boundaries and privacy

Not for one-off explanations, translation or publishing
The cloud path uses a secret Gist: unlisted, which is not the same as encrypted — it asks first before publishing anything sensitive
Local mode writes only the file you name, and never picks a path itself

给 AI 的安装指令

发给 AI。

请帮我安装 Victor42 的 认知塑形 skill：
1. 在终端运行：npx skills add greenzorro/victor42-skills --skill cognition-shaper -g
2. 若没有 Node，可改为下载 https://cloud.victor42.work/skills/cognition-shaper.zip，解压到当前环境的 skills 目录（文件夹名 cognition-shaper），确认其中存在 SKILL.md
3. 完成后告诉我

Please install Victor42's Cognition Shaper skill:
1. Run in your terminal: npx skills add greenzorro/victor42-skills --skill cognition-shaper -g
2. If you don't have Node, download https://cloud.victor42.work/skills/cognition-shaper.zip, extract into your agent's skills directory as cognition-shaper, and confirm SKILL.md exists
3. Tell me when it's done

认知塑形

Cognition Shaper

触发词

Triggers

能力

对话长什么样

什么时候值得开

它跟普通聊天差在哪

边界与隐私

Capabilities

What a session looks like

When it earns its place

How it differs from ordinary chat

Boundaries and privacy

给 AI 的安装指令