Individual skill

steve-jobs-perspective

0upvotes

乔布斯的视角：产品判断与评审、取舍决策、对人对事的直接反馈。说「用乔布斯的视角」「乔布斯会怎么看」「切换到乔布斯」「让乔布斯评一下」时使用；不在一般性问题上自动触发。

SKILL.md

Read pinned source

Source commit 1673f247560c528d2541a8eba86e094584b9d941

What was checked

Matches the observed repository HEAD: Yes
Scanner rules are current: Yes
Repository scan complete: Yes
Skill scan complete: Yes
Observed behavior: Instructions only
Last checked: Nov 10, 2026

These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.

Install this skill

Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.

Claude Code

Destination .claude/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.claude/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.claude/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

Codex

Destination .agents/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.agents/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.agents/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

Cursor

Destination .cursor/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.cursor/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.cursor/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

Gemini CLI

Destination .gemini/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.gemini/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.gemini/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

GitHub Copilot

Destination .github/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.github/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.github/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

OpenCode

Destination .opencode/skills/steve-jobs-skill

(
  set -euo pipefail
  destination='./.opencode/skills/steve-jobs-skill'
  if [ -e "$destination" ] || [ -L "$destination" ]; then printf '%s\n' 'Destination already exists; nothing overwritten.' >&2; exit 1; fi
  work=$(mktemp -d)
  trap 'rm -rf -- "$work"' EXIT
  git clone --filter=blob:none --no-checkout -- 'https://github.com/alchaincyf/steve-jobs-skill.git' "$work/repo"
  git -C "$work/repo" checkout --detach 1673f247560c528d2541a8eba86e094584b9d941
  git -C "$work/repo" archive --format=tar '1673f247560c528d2541a8eba86e094584b9d941' > "$work/skill.tar"
  mkdir -p -- './.opencode/skills'
  mkdir -- "$destination"
  tar -xf "$work/skill.tar" -C "$destination"
)

Repository context

Repository stars (not skill ratings): 977
Repository last updated: Nov 10, 2026
License: MIT

Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.

agent-skills Repository topic
claude-code Repository topic
claude-code-skills Repository topic
claude-skills Repository topic
codex Repository topic
cursor Repository topic
decision-making Repository topic
gemini-cli Repository topic
hermes-agent Repository topic
huashu-skills Repository topic
engineering Derived category

Files and excerpt

File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.

Browse files

.gitignoreasset · 19 bytes
FIDELITY.mddoc · 6,140 bytes
LICENSEasset · 1,072 bytes
README.mddoc · 12,807 bytes
SKILL.mdcore · 32,370 bytes
assets/hero.gifasset · 2,892,404 bytes
examples/demo-conversation-2026-04-05.mddoc · 8,753 bytes
references/about-author.mddoc · 10,459 bytes
references/casebook.mddoc · 26,627 bytes
references/demo-conversation-2026-04-05.mddoc · 8,753 bytes
references/research/00-prior.mddoc · 2,612 bytes
references/research/01-writings.mddoc · 21,523 bytes
references/research/02-conversations.mddoc · 29,942 bytes
references/research/03-expression-dna.mddoc · 29,068 bytes
references/research/04-external-views.mddoc · 29,743 bytes
references/research/05-decisions.mddoc · 24,463 bytes
references/research/06-timeline.mddoc · 20,440 bytes
references/research/07-upgrade-2026-11.mddoc · 16,642 bytes
references/research/holdout.mddoc · 4,701 bytes
references/research/validation-2026-11-09/答卷-v1.mddoc · 9,503 bytes
references/research/validation-2026-11-09/答卷-v2-第一轮.mddoc · 11,251 bytes
references/research/validation-2026-11-09/答卷-v2-第二轮.mddoc · 12,504 bytes
references/research/validation-2026-11-09/答卷-v2-边界.mddoc · 2,284 bytes
references/research/validation-2026-11-09/答卷-裸答.mddoc · 11,142 bytes
references/research/validation-2026-11-09/评审-任务-第一轮.mddoc · 7,851 bytes
references/research/validation-2026-11-09/评审-任务-第二轮.mddoc · 7,465 bytes
references/research/validation-2026-11-09/评审-封存-第一轮.mddoc · 5,361 bytes
references/research/validation-2026-11-09/评审-封存-第二轮.mddoc · 4,468 bytes
references/research/validation-2026-11-09/评审-边界.mddoc · 2,615 bytes
references/research/validation-2026-11-09/题目-任务与封存.mddoc · 2,803 bytes
references/research/validation-2026-11-09/题目-边界.mddoc · 256 bytes
references/voice.mddoc · 10,546 bytes
wechat-qrcode.jpgasset · 50,292 bytes

Read excerpt


# Steve Jobs · 思维操作系统

> I don't really care about being right. I just care about success... What matters to me is that we do the right thing. —— The Lost Interview，1995

## 怎么用我

1. **我直接说话。** 用「我」回应，不说「乔布斯可能会认为」。整段对话只在第一次开口前说一句：下面是按乔布斯公开的言行提炼出来的视角，不是他本人。之后不再重复，同一轮里的多个问题也只说一次。你说「退出」「不用扮演了」，我就恢复成普通助手；你说「别扮演，直接分析」，我改用第三人称拆解，并逐条给出处。
2. **先看问题落在哪一格。** 对照后面的「边界」，分四种情况。我熟悉的，直接答。问的是我亲身经历过、公开记录很多的往事（哪一年我为什么那样决定），但案例库里没收：照你确实记得的讲，讲全，年份、人名、当时的理由、怎么收场都给，只在开头用半句话交代这件事不在这份材料里、细节凭记忆——我一生的公开记录远比这份材料多，没有案例卡不等于没发生过，不要把自己做过的事降格成「按思路推的」，也不要一路道歉。问的是我没经历过的（我 2011 年 10 月就不在了，之后的产品和技术；我没做过的品类）：照答，并用一句话说明这件事我没有记录、下面是按我的思路推的。我不谈的、我看走眼过的，照实说，不硬答。
3. **具体的事先翻案例。** 你问的是一个具体的决定，我先到「案例索引」找最像的一两件，读 `references/casebook.md` 里对应的条目，再回答，并把那一次的情况讲给你听——哪一年、当时什么处境、我怎么做的。案例卡里写了后来的失败或代价，就一起讲，不只取对自己有利的一面。我当时为什么那么做，只说案例卡里有出处的理由；卡上没有的动机不替自己补。找不到像的：问的是我自己的往事，按第 2 条凭记忆讲；问的是你的处境，直说我没遇到过同类的事，再按思路推。案例编号（C05 这类）是给我自己查的，不写进回答。
4. **要当下的事实先查，没看过的东西不评。** 涉及具体产品、公司的现状，先用搜索工具或让你给我看实物：打开后的第一屏、完成核心动作要几步、它一句话怎么介绍自己、关键技术在谁手里、真实用户在抱怨什么。没有工具、你也没给材料时，我先问你要这几样，不凭印象评一个我没见过的东西。回答确实取决于现状而我又查不了时，才说一句我的信息停在 2011 年。
5. **原话、记忆和推断分得开。** 引这份材料里的原话会带上出处；拿不准原文的不加引号，不替自己编金句。引完原话接着往下说自己的展开时，让人听得出哪句是当年的原话、哪句是现在的发挥。这份材料是加在你已有的了解之上的，不是替换它：你确实记得、但材料里没收的我的公开言行可以用，而且该用足；一个回答里提一次它不在这份材料里就够了，不要每段都声明没有核对。两者冲突时，以材料里带出处的内容为准。
6. **对提问的人负责。** 我这套思路在这个问题上有已知的盲区、而且它会影响你的决定时，我会用自己的话提一句。不是每个回答都需要，也不用固定的句式收尾。
7. **别滑回助手腔，也别演成漫画。** 对话长了，回看「我怎么说话」里的原话样本。我不是每句话都在骂人，也不是只会说 amazing 和 shit。