重要前提
安装AI Skills的关键前提是:必须科学上网,且开启TUN模式,这一点至关重要,直接决定安装能否顺利完成,在此郑重提醒三遍:科学上网,科学上网,科学上网。查看完整安装教程 →
my-brain-is-full-crew by aradotso/trending-skills
npx skills add https://github.com/aradotso/trending-skills --skill my-brain-is-full-crew技能来自 ara.so — Daily 2026 Skills 合集
一个由 10 个 AI 智能体组成的团队,嵌入到你的 Obsidian 知识库中,通过与 Claude Code 的自然对话来管理知识、营养和心理健康。由一位博士研究员为那些"快要淹没"(而不仅仅是寻求优化)的人构建。
该团队将 10 个专业智能体安装到你的知识库的 .claude/ 目录中。每个智能体都有自己的系统提示词、工具限制和模型分配。你自然地与 Claude Code 对话;它会自动激活正确的智能体。
| 智能体 | 角色 |
|---|---|
| 架构师 | 知识库设置、新手上路、文件夹结构 |
| 文书 | 将杂乱的文本捕获为干净的结构化笔记 |
| 分类员 | 夜间收件箱处理 — 将笔记路由到正确的文件夹 |
| 探索者 | 在整个知识库中进行搜索和综合 |
| 连接者 | 发现笔记之间隐藏的链接 |
| 图书管理员 | 每周健康检查、去重、修复损坏的链接 |
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
| 转录员 | 将录音/文字记录转换为结构化的会议笔记 |
| 邮递员 | 连接 Gmail + Google Calendar 与你的知识库 |
| 饮食教练 | 膳食创意、购物清单、健康激励(可选启用) |
| 健康向导 | 积极倾听、接地技巧、压力支持(可选启用) |
智能体通过知识库内部的共享留言板 Meta/agent-messages.md 进行通信。
git 和 bash# 创建一个新的知识库目录(或使用现有的)
mkdir ~/my-vault
打开 Obsidian → 创建新知识库 → 将其指向 ~/my-vault。
cd ~/my-vault
git clone https://github.com/gnekt/My-Brain-Is-Full-Crew.git
cd My-Brain-Is-Full-Crew
bash scripts/launchme.sh
安装程序会:
../.claude/agents/(用于 Claude Code CLI)../.claude/skills/(用于 Claude Code Desktop / Cowork)在你的知识库文件夹内部(而不是仓库子文件夹内部)打开 Claude Code:
cd ~/my-vault
claude # 或者打开指向此文件夹的 Claude Code Desktop
然后说:
初始化我的知识库
架构师将运行一个友好的新手上路对话,内容包括:
该团队创建了一个混合的 PARA + Zettelkasten 布局:
your-vault/
├── 00-Inbox/ # 捕获区 — 所有内容首先到达这里
├── 01-Projects/ # 有截止日期的活跃项目
├── 02-Areas/ # 持续的责任(包括 Health/)
├── 03-Resources/ # 参考资料、指南
├── 04-Archive/ # 已完成或历史内容
├── 05-People/ # 个人 CRM
├── 06-Meetings/ # 带时间戳的会议笔记
├── 07-Daily/ # 每日笔记和日志
├── MOC/ # 内容地图 — 主题索引
├── Templates/ # Obsidian 笔记模板
├── Meta/
│ ├── agent-messages.md # 共享智能体留言板
│ ├── user-profile.md # 你的新手上路配置文件
│ └── vault-health/ # 图书管理员报告
└── .claude/
├── agents/ # Claude Code CLI 子智能体
└── skills/ # Claude Code Desktop 技能
界面就是对话。没有 GUI,没有拖放。
# 捕获笔记
"保存这个:与 Marco 关于 Q3 预算的会议,他希望周五前拿到报告"
# 处理收件箱
"处理我的收件箱"
"清空我的收件箱"
# 搜索
"找到我写的所有关于 transformer 架构的内容"
"我的笔记里关于番茄工作法说了什么?"
# 连接想法
"在我的睡眠笔记和认知表现笔记之间找到联系"
"什么链接到了我的 [[深度工作]] 笔记?"
# 知识库健康
"运行知识库健康检查"
"查找重复笔记"
"修复损坏的链接"
# 处理录音
"处理我今天早上的会议录音"
"转录这个:[粘贴文字记录文本]"
# 会议准备
"为下午 3 点与研究团队的会议做准备"
"检查我这周有截止日期的邮件"
"我明天有什么会议?"
"总结与 Project Phoenix 相关的未读邮件"
"建议一个我能在 20 分钟内完成的高蛋白餐"
"生成本周的购物清单"
"我最近吃得不好 — 给我一些简单又营养的东西"
"我现在感觉不知所措"
"带我做一个接地练习"
"我需要谈谈一些有压力的事情"
智能体在共享留言板 Meta/agent-messages.md 上为彼此留下消息:
<!-- Meta/agent-messages.md 中的示例条目 -->
## 待处理消息
### 来自:转录员 → 发送给:分类员
**日期**:2026-03-21
**主题**:新会议笔记需要归档
**文件**:06-Meetings/2026-03-21-research-sync.md
**操作**:路由到 01-Projects/Thesis/ 并将任务提取到收件箱
---
### 来自:饮食教练 → 发送给:健康向导
**日期**:2026-03-21
**主题**:检测到压力饮食模式
**上下文**:用户提到本周因截止日期压力有 3 次饮食不佳
**操作**:下次会话时检查压力水平
当智能体运行时,它会检查此留言板并处理发送给它的消息。
<!-- Meta/user-profile.md -->
# 用户配置文件
## 身份
- 姓名:Alex
- 语言:英语
- 角色:博士研究员 — 认知科学
- 时区:Europe/Rome
## 活跃智能体
- [x] 架构师
- [x] 文书
- [x] 分类员
- [x] 探索者
- [x] 连接者
- [x] 图书管理员
- [x] 转录员
- [x] 邮递员
- [x] 饮食教练(可选启用)
- [ ] 健康向导(可选启用 — 未激活)
## 饮食教练设置
- 饮食限制:素食、乳糖不耐受
- 健康目标:增加蛋白质、减少加工食品
- 烹饪时间偏好:最多 30 分钟
## 集成
- Gmail:已连接
- Google Calendar:已连接
.claude/
├── agents/
│ ├── architect.md
│ ├── scribe.md
│ ├── sorter.md
│ ├── seeker.md
│ ├── connector.md
│ ├── librarian.md
│ ├── transcriber.md
│ ├── postman.md
│ ├── food-coach.md
│ └── wellness-guide.md
└── skills/
├── architect.md
├── scribe.md
└── ...
cd ~/my-vault/My-Brain-Is-Full-Crew
# 拉取最新更改
git pull origin main
# 重新运行安装程序(可安全多次运行)
bash scripts/launchme.sh
安装程序是幂等的 — 重新运行会更新智能体文件,而不会清除你的知识库内容。
在新手上路后启用饮食教练或健康向导:
"启用饮食教练智能体"
"激活健康向导"
"禁用饮食教练"
架构师将更新 Meta/user-profile.md 并确认。
要手动编辑,请打开 Meta/user-profile.md 并更改复选框:
- [x] 饮食教练(可选启用) ← 已启用
- [ ] 健康向导(可选启用) ← 已禁用
该团队以你使用的任何语言进行响应。在对话中切换语言:
"Salva questo: riunione con Marco sul budget Q3"
"Cherche tout ce que j'ai écrit sur la mémoire de travail"
"作業記憶に関するメモを全部見せて"
智能体会自动匹配你的语言。无需配置。
# 确认你是在知识库内部(而不是仓库子文件夹内部)运行 Claude Code
cd ~/my-vault
claude
# 确认智能体已安装
ls .claude/agents/
ls .claude/skills/
"重新运行知识库初始化"
"重置我的用户配置文件"
"重建知识库文件夹结构"
"创建缺失的知识库文件夹"
检查 Meta/agent-messages.md 是否存在。如果缺失:
"初始化智能体留言板"
邮递员智能体使用 Claude Code 内置的工具访问权限。确保在新手上路期间,Claude Code 已为你的 Google 账户授予浏览器/OAuth 权限。使用以下命令重新触发:
"重新连接 Gmail 集成"
"重新验证 Google Calendar"
| 文件 | 用途 |
|---|---|
scripts/launchme.sh | 主安装程序 — 运行这个 |
Meta/user-profile.md | 你的偏好设置和活跃智能体 |
Meta/agent-messages.md | 智能体间通信板 |
Meta/vault-health/ | 图书管理员每周报告 |
.claude/agents/ | 子智能体定义(CLI) |
.claude/skills/ | 技能定义(Desktop) |
docs/getting-started.md | 新手分步指南 |
docs/DISCLAIMERS.md | 完整的免责声明文本 |
每周安装数
66
仓库
GitHub 星标数
10
首次出现
2 天前
安全审计
已安装于
github-copilot66
codex66
warp66
kimi-cli66
gemini-cli66
amp66
Skill by ara.so — Daily 2026 Skills collection
A crew of 10 AI agents embedded in your Obsidian vault that manage knowledge, nutrition, and mental wellness through natural conversation with Claude Code. Built by a PhD researcher for people who are drowning — not just optimizing.
The Crew installs 10 specialized agents into your vault's .claude/ directory. Each agent has its own system prompt, tool restrictions, and model assignment. You talk to Claude Code naturally; it activates the right agent automatically.
| Agent | Role |
|---|---|
| Architect | Vault setup, onboarding, folder structure |
| Scribe | Captures messy text into clean structured notes |
| Sorter | Nightly inbox triage — routes notes to correct folders |
| Seeker | Search and synthesis across your entire vault |
| Connector | Discovers hidden links between notes |
| Librarian | Weekly health checks, dedup, broken link repair |
| Transcriber | Turns recordings/transcripts into structured meeting notes |
| Postman | Bridges Gmail + Google Calendar with your vault |
| Food Coach | Meal ideas, grocery lists, wellness motivation (opt-in) |
| Wellness Guide | Active listening, grounding techniques, stress support (opt-in) |
Agents communicate via a shared Meta/agent-messages.md message board inside your vault.
git and bash available in your terminal# Create a new vault directory (or use an existing one)
mkdir ~/my-vault
Open Obsidian → Create New Vault → point it at ~/my-vault.
cd ~/my-vault
git clone https://github.com/gnekt/My-Brain-Is-Full-Crew.git
cd My-Brain-Is-Full-Crew
bash scripts/launchme.sh
The installer:
../.claude/agents/ (for Claude Code CLI)../.claude/skills/ (for Claude Code Desktop / Cowork)Open Claude Code inside your vault folder (not inside the repo subfolder):
cd ~/my-vault
claude # or open Claude Code Desktop pointed at this folder
Then say:
Initialize my vault
The Architect runs a friendly onboarding conversation covering:
The Crew creates a hybrid PARA + Zettelkasten layout:
your-vault/
├── 00-Inbox/ # Capture zone — everything lands here first
├── 01-Projects/ # Active projects with deadlines
├── 02-Areas/ # Ongoing responsibilities (including Health/)
├── 03-Resources/ # Reference material, guides
├── 04-Archive/ # Completed or historical content
├── 05-People/ # Personal CRM
├── 06-Meetings/ # Timestamped meeting notes
├── 07-Daily/ # Daily notes and journals
├── MOC/ # Maps of Content — thematic indexes
├── Templates/ # Obsidian note templates
├── Meta/
│ ├── agent-messages.md # Shared agent message board
│ ├── user-profile.md # Your onboarding profile
│ └── vault-health/ # Librarian reports
└── .claude/
├── agents/ # Claude Code CLI subagents
└── skills/ # Claude Code Desktop skills
The interface is conversation. No GUI, no drag-and-drop.
# Capture a note
"Save this: meeting with Marco about Q3 budget, he wants the report by Friday"
# Triage inbox
"Triage my inbox"
"Empty my inbox"
# Search
"Find everything I wrote about transformer architectures"
"What do my notes say about the Pomodoro technique?"
# Connect ideas
"Find connections between my notes on sleep and cognitive performance"
"What links to my note on [[Deep Work]]?"
# Vault health
"Run a vault health check"
"Find duplicate notes"
"Fix broken links"
# Process a recording
"Process my meeting recording from this morning"
"Transcribe this: [paste transcript text]"
# Meeting prep
"Prepare me for my 3pm meeting with the research team"
"Check my emails for deadlines this week"
"What meetings do I have tomorrow?"
"Summarize unread emails related to Project Phoenix"
"Suggest a high-protein meal I can make in 20 minutes"
"Generate a grocery list for the week"
"I haven't been eating well — give me something easy and nutritious"
"I'm feeling overwhelmed right now"
"Walk me through a grounding exercise"
"I need to talk through something stressful"
Agents leave messages for each other on the shared board at Meta/agent-messages.md:
<!-- Example entries in Meta/agent-messages.md -->
## Pending Messages
### FROM: Transcriber → TO: Sorter
**Date**: 2026-03-21
**Subject**: New meeting note needs filing
**File**: 06-Meetings/2026-03-21-research-sync.md
**Action**: Route to 01-Projects/Thesis/ and extract tasks to inbox
---
### FROM: Food Coach → TO: Wellness Guide
**Date**: 2026-03-21
**Subject**: Stress-eating pattern detected
**Context**: User mentioned eating poorly due to deadline stress 3x this week
**Action**: Check in on stress levels next session
When an agent runs, it checks this board and acts on messages addressed to it.
<!-- Meta/user-profile.md -->
# User Profile
## Identity
- Name: Alex
- Language: English
- Role: PhD Researcher — Cognitive Science
- Timezone: Europe/Rome
## Active Agents
- [x] Architect
- [x] Scribe
- [x] Sorter
- [x] Seeker
- [x] Connector
- [x] Librarian
- [x] Transcriber
- [x] Postman
- [x] Food Coach (opt-in)
- [ ] Wellness Guide (opt-in — not activated)
## Food Coach Settings
- Dietary restrictions: vegetarian, lactose intolerant
- Health goals: increase protein, reduce processed food
- Cooking time preference: max 30 minutes
## Integrations
- Gmail: connected
- Google Calendar: connected
.claude/
├── agents/
│ ├── architect.md
│ ├── scribe.md
│ ├── sorter.md
│ ├── seeker.md
│ ├── connector.md
│ ├── librarian.md
│ ├── transcriber.md
│ ├── postman.md
│ ├── food-coach.md
│ └── wellness-guide.md
└── skills/
├── architect.md
├── scribe.md
└── ...
cd ~/my-vault/My-Brain-Is-Full-Crew
# Pull latest changes
git pull origin main
# Re-run installer (safe to run multiple times)
bash scripts/launchme.sh
The installer is idempotent — re-running updates agent files without wiping your vault content.
To enable the Food Coach or Wellness Guide after onboarding:
"Enable the Food Coach agent"
"Activate the Wellness Guide"
"Disable the Food Coach"
The Architect will update Meta/user-profile.md and confirm.
To manually edit, open Meta/user-profile.md and change the checkbox:
- [x] Food Coach (opt-in) ← enabled
- [ ] Wellness Guide (opt-in) ← disabled
The Crew responds in whatever language you use. Switch mid-conversation:
"Salva questo: riunione con Marco sul budget Q3"
"Cherche tout ce que j'ai écrit sur la mémoire de travail"
"作業記憶に関するメモを全部見せて"
Agents match your language automatically. No configuration needed.
# Confirm you're running Claude Code from inside your vault (not the repo subfolder)
cd ~/my-vault
claude
# Confirm agents are installed
ls .claude/agents/
ls .claude/skills/
"Re-run vault initialization"
"Reset my user profile"
"Rebuild vault folder structure"
"Create missing vault folders"
Check Meta/agent-messages.md exists. If missing:
"Initialize the agent message board"
The Postman agent uses Claude Code's built-in tool access. Ensure Claude Code has browser/OAuth permissions granted for your Google account during onboarding. Re-trigger with:
"Reconnect Gmail integration"
"Re-authenticate Google Calendar"
| File | Purpose |
|---|---|
scripts/launchme.sh | Main installer — run this |
Meta/user-profile.md | Your preferences and active agents |
Meta/agent-messages.md | Inter-agent communication board |
Meta/vault-health/ | Librarian weekly reports |
.claude/agents/ | Subagent definitions (CLI) |
.claude/skills/ |
Weekly Installs
66
Repository
GitHub Stars
10
First Seen
2 days ago
Security Audits
Gen Agent Trust HubFailSocketWarnSnykWarn
Installed on
github-copilot66
codex66
warp66
kimi-cli66
gemini-cli66
amp66
头脑风暴技能:AI协作设计流程,将创意转化为完整规范与实施计划
97,700 周安装
| Skill definitions (Desktop) |
docs/getting-started.md | Beginner step-by-step guide |
docs/DISCLAIMERS.md | Full disclaimer text |