recording-knowledge

在 AI 辅助开发过程中，自动将特性、决策、踩坑记录到项目本地的 docs/kb/。

Automatically record features, decisions, and pitfalls during AI-assisted development — zero config, zero dependencies, one skill.

---

English · 中文

---

English

Installation

npx skills add https://github.com/CCass/recording-knowledge

Or browse at skills.sh.

What it does

When installed, your AI agent will automatically:

• Initialize — Create docs/kb/ skeleton when entering a project without one
• Record — Proactively save features, technical decisions, and pitfalls as Markdown files
• Index — Maintain INDEX.md automatically
• Review — Check for unrecorded knowledge at session end
• Question — Before designing, ask WHAT→WHY→ACTUAL NEED to prevent over-engineering

Knowledge base structure

docs/kb/
├── INDEX.md           ← Auto-maintained index
├── TEMPLATE.md        ← Record template
├── features/    (f-)  ← Completed features
├── decisions/   (d-)  ← Technical decisions
├── pitfalls/    (p-)  ← Lessons learned
└── architecture/ (a-) ← Architecture snapshots

How it works

Trigger	Type	Example
Agent detects a completed feature	Feature (f-)	"User registration implemented"
Agent detects a technical decision	Decision (d-)	"Chose A over B framework"
Agent detects a bug fix / pitfall	Pitfall (p-)	"MySQL JSON index failure"
User says "记录一下" / "record this"	Any	Manual trigger

Who is this for

• Solo developers using AI coding tools
• Teams wanting project-level knowledge preservation
• Anyone tired of losing context between AI sessions

License

MIT

---

中文

安装

npx skills add https://github.com/CCass/recording-knowledge

或在 skills.sh 浏览。

功能

安装后，AI Agent 会自动：

• 初始化 — 进入无知识库的项目时自动创建 docs/kb/ 骨架
• 记录 — 主动将特性、决策、踩坑保存为 Markdown 文件
• 索引 — 自动维护 INDEX.md
• 检查 — 会话结束前检查是否有遗漏未记录的知识
• 追问 — 设计前先问 WHAT→WHY→ACTUAL NEED，防过度设计

知识库结构

docs/kb/
├── INDEX.md           ← 主索引（自动维护）
├── TEMPLATE.md        ← 记录模板
├── features/    (f-)  ← 已开发特性
├── decisions/   (d-)  ← 技术决策
├── pitfalls/    (p-)  ← 踩坑记录
└── architecture/ (a-) ← 架构快照

触发方式

场景	类型	例子
Agent 识别到完成了一个功能	Feature (f-)	"用户注册功能已实现"
Agent 识别到做了技术选型	Decision (d-)	"选了 A 框架而非 B 框架"
Agent 识别到踩坑修复	Pitfall (p-)	"MySQL JSON 索引失效"
用户说"记录一下"	任意	手动触发

适用人群

• 使用 AI 编程的个人开发者
• 希望项目级知识沉淀的团队
• 厌倦了 AI 会话之间丢失上下文的任何人

协议

MIT