双重(陀思妥耶夫斯基参考)特工本地记忆系统
概览
为Claude代码、光标及其他AI工具提供持久上下文。跳过数据库、MCP服务器、简单markdown文件的嵌入和老式git。 这是一个公开模板。一旦你克隆并添加了笔记,不要将个人版本推送到公共仓库。你的内存将包含秘密(包括暗黑的个人秘密和安全风险秘密)、个人信息以及你的工作风格。保持本地或推送到新的私有仓库。- 公共仓库的分支不能设为私密。请使用git克隆(见下方快速入门)。 AI助理在每次会议之间都会忘记所有内容。你的工程决策(我认为尤其重要)、编码模式、项目背景。 Double 通过你控制的有序 markdown 文件解决这个问题。把它当作你的 AI 助手的第二个大脑。 以陀思妥耶夫斯基的小说《双重人》命名——这是你的第二个自己。不仅是工程和研究,还有商业想法、个人笔记、爱好、工作流程偏好。
概述(AI 翻译)
为Claude代码、光标及其他AI工具提供持久上下文。跳过数据库、MCP服务器、简单markdown文件的嵌入和老式git。
警告: 这是一个公开模板。一旦你克隆并添加了笔记,千万不要将个人版本推送到公共仓库。你的内存将包含秘密(包括黑暗的个人秘密和安全风险的秘密)、个人信息以及你的工作风格。保持本地或推送到新的私人仓库。
不要使用 GitHub 的“分叉”按钮 - 公共仓库的分叉不能设为私密。请使用
git clone(见下方快速入门)。
什么是双重?
AI助理在每次会议之间都会忘记所有内容。你的工程决策(我认为尤其重要)、编码模式、项目背景。
Double 通过你控制的有序 markdown 文件解决这个问题。把它当作你的 AI 助手的第二个大脑。
为什么叫“双重”? 以陀思妥耶夫斯基的小说《双重》命名——这是你的第二个自己。不仅仅是工程和研究,还有商业想法、个人笔记、爱好、工作流程偏好。
主要特色
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
Double: External Personal Memory for AI
Persistent context for Claude Code, Cursor, and other AI tools. Skips the database, MCP servers, embeddings for simple markdown files and good ol’ git.
WARNING: This is a PUBLIC template. Once you clone it and add your notes, DO NOT push your personal version to a public repo. Your memory will contain secrets (both dark personal secrets and security risk secrets), personal information and your working style. Keep it local or push to a NEW private repo.
DO NOT use GitHub’s “Fork” button - forks of public repos cannot be made private. Use
git cloneinstead (see Quick Start below).
What is Double?
AI assistants forget everything between sessions. Your engineering decisions (which I think are especially important), coding patterns, project context.
Double solves this with organised markdown files that you control. Think of it as a second brain for your AI assistant.
Why “Double”? Named after Dostoevsky’s novel “The Double” - it’s your second self. Not just engineering and research, but business ideas, personal notes, hobbies, workflow preferences.
Key Features
- Smart routing:
/synconly loads files relevant to your work - Python session won’t load iOS context - Concise capture:
/handoffwrites terse, reusable insights (not verbose essays) - Context efficient: Compressed memory files reduce token overhead by ~65% while keeping essential knowledge
- No dependencies: Just markdown files and git
Make It Yours
This is a template, not a rigid system. The folder structure, commands, and workflows here are a starting point based on one person’s working style. You should heavily customise it:
- Rename folders to match your mental model
- Delete categories you don’t need (maybe you don’t track business ideas or research)
- Add new ones that fit your work (design/, writing/, client-work/, whatever)
- Modify commands to match your workflow
- Change the rules in
meta/to fit your style
The only critical parts are:
- Organised markdown files you control
- A capture → process → load workflow
- Using absolute paths (
~/double/) in commands
Everything else? Change it.
Folder Structure
~/double/
├── business/ # Ideas, metrics, product decisions
├── engineering/ # Technical patterns, best practices, lessons learned
├── tasks/ # Active work, backlog, waiting
├── projects/ # Project-specific context and decisions
├── research/ # Intelligence on companies, people, markets
├── personal/ # Writing style, preferences, workflows
├── meta/ # System rules, processed entries, changelog
└── .claude/
├── commands/ # Workflow commands that load context
└── skills/ # Forked skills (tangent, etc.)
Why ~/double/ in Your Home Directory?
推荐工具
换一个关键词,或者移除筛选条件。
安装
npx skillfish add ossa-ma/double