123 Flow — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited 123 Flow (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
Aggregate score unchanged between these scans.
The primary manifest — the file an agent reads to learn what this artifact does.
<div align="center">
`1` 干到底 · `2` 往死里审 · `3` 收尾上线 `1` ship it · `2` tear it apart · `3` go live
</div>
给 AI 装一个方向盘: 打1让它干、打2让它审、打3让它上线。 一个字符切换工作流阶段 —— 不用每次组织一长串话, 而且逼 AI 别敷衍。
Give your AI a steering wheel: hit1to execute,2to review,3to ship. One keystroke switches workflow phase — no more typing full sentences, and it stops the AI from cutting corners.
flowchart LR
A(["你打一个数字<br/>You hit a key"]) --> K1["1 = 干 / Do"]
A --> K2["2 = 审 / Review"]
A --> K3["3 = 完工 / Ship"]
K1 --> D1["第一性思考 · 费曼回答<br/>复盘 + 浏览器实测<br/>部署上线"]
K2 --> D2["0→100 + 100→0 双向<br/>逆向反驳<br/>13 / 21 / 18 项清单"]
K3 --> D3["FP + 费曼 → 复盘<br/>→ 真机找 BUG<br/>→ 部署 exit 0"]
style K1 fill:#7C3AED,color:#fff
style K2 fill:#DC2626,color:#fff
style K3 fill:#16A34A,color:#fff你跟 AI 写代码, 是不是每天重复打这几句:
一天几十遍。手累, 而且 AI 常敷衍 —— 嘴上说"搞定了", 实际没测、没上线、藏着雷。
Every day you retype the same three things: "keep going", "wait, check for bugs", "is it actually shipped?". Tiring — and the AI keeps claiming "done" when it isn't.
三件事各绑一个数字键。打一个字符, AI 进入对应的完整工作流:
| 键 | 一句话 | 背后跑的是 |
|---|---|---|
| `1` | 干到底 | 全速执行 + 第一性思考 + 费曼回答 + 自动复盘 + 浏览器实测 + 部署上线 |
| `2` | 往死里审 | 0→100 + 100→0 双向复盘 + 逆向反驳 + 13 / 21 / 18 项分领域清单 |
| `3` | 收尾上线 | 第一性+费曼 → 复盘 → 真机实测找 BUG → 部署验证 exit 0 |
核心: 打1不是敷衍地继续, 是带着一整套铁律继续。打2是逼 AI 当自己的对手。打3是真的开浏览器走一遍 + 确认线上能用。
对号入座 —— 凡是这些情况, 你只需打一个数字键:
| 你遇到这种情况 | 打 | 会发生什么 |
|---|---|---|
| AI 给了方案 / 建议, 你认可, 懒得再打一长串"那就这样开始干" | 1 | 直接执行到底, 自带全套铁律 (第一性思考 → 复盘 → 浏览器实测 → 上线) |
| AI 说"做完了 / 修好了", 你将信将疑 | 2 | 逼它当自己的对手, 双向复盘 + 揪出它没测的边界 |
| AI 凭记忆瞎写, 你担心 API / 用法用错 | 2 | 复盘清单里"查证过"那条, 逼它真去搜证实 |
| 功能开发完准备上线, 怕它没真部署 / 没测手机 | 3 | 强制真机实测 + 部署验证, exit 0 才放行 |
| 深夜赶工, 懒得组织语言 | 1 2 3 | 三个键搞定"干 → 审 → 上线"整个节奏 |
一句话: 凡是你想对 AI 说"继续 / 你确定吗 / 上线了吗", 都能压成一个数字键。 Whenever you'd type "keep going" / "are you sure?" / "is it live?" — compress it into one keystroke.
2 复盘是灵魂 / The Soul is 2会写代码的 AI 满地都是, 但能回头审自己、戳穿"虚假完成"的 AI 很少。 AI 默认讨好你 —— 你说改它就改, 你不问它就说"好了"。2 强制它反向找茬:
Any AI can write code. Few can turn around and tear their own work apart. `2` forces the AI to be its own adversary — bidirectional review, reverse rebuttal, and a domain-specific checklist of up to 21 sharp questions.
完整清单 (逐条大白话解释) 见 skills/123-flow/SKILL.md。
A. Plugin 一键装 (推荐 / recommended):
/plugin marketplace add LouisLin0723/123-flow
/plugin install 123-flow@123-flowB. 手动 clone / manual:
git clone https://github.com/LouisLin0723/123-flow.git
cp -r 123-flow/skills/123-flow ~/.claude/skills/重启 Claude Code 生效。完整方法论 → skills/123-flow/SKILL.md。
2 强制反向找茬, 戳穿"虚假的完成"。这是灵魂。1 往前冲、2 回头审、3 真上线, 三键拼成一个完整的 AI 协作闭环, 不留尾巴。这套 1/2/3 是作者的工作流约定。Fork 它, 改 SKILL.md 里的触发词、检查清单、部署命令, 换成你自己的节奏。
它不绑定任何具体项目或框架 —— 是一套可迁移的工作流骨架。 Project- and framework-agnostic. A portable workflow skeleton, not a config dump.
MIT © Louis Lin
<div align="center">
这套工作流帮你省了脑子? 点个 ⭐ Star 收藏一下 ❤️ Saved you some brain cycles? Drop a ⭐ — it helps others find it.
</div>
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.