AI学习日志技能：智能代理助你捕获、回顾与应用知识，告别遗忘

Learning Journal by theagentledger/agent-skills

GitHub

安装命令

npx skills add https://github.com/theagentledger/agent-skills --skill 'Learning Journal'

AI/机器学习生产力知识管理

🇨🇳中文介绍

学习日志

作者：The Agent Ledger — 为那些消费多于留存的人设计的 AI 原生学习系统。

您的智能体追踪您正在学习的内容，提示您回顾关键概念，并确保知识真正扎根 — 并得到应用。

问题所在

您每周阅读 30 篇文章。您收听播客。您参加课程。但 3 个月后，您却无法回忆起那本“改变一切”的书籍中的关键见解。大多数人的消费与留存比率非常糟糕。

这项技能将您的 AI 智能体转变为一个学习伙伴，能够随时间推移捕获、组织、回顾和连接知识。

设置

步骤 1：创建状态文件

在您的工作区中创建 learning/learning-state.md：

# 学习状态

## 活跃学习轨道
| 轨道 | 类型 | 来源 | 开始日期 | 进度 | 优先级 |
|-------|------|--------|---------|----------|----------|
| [主题] | 书籍 | [书名 作者] | YYYY-MM-DD | 35% | 高 |
| [主题] | 课程 | [课程名称 — 平台] | YYYY-MM-DD | 60% | 中 |
| [主题] | 自主学习 | [描述] | YYYY-MM-DD | 进行中 | 低 |

## 学习目标（本季度）
1. [完成 X] — 截止日期：[日期]
2. [充分理解 Y 以便 Z] — 无截止日期，但有优先级
3. [建立 Z 技能] — 可衡量：[证明标准]

## 回顾计划
- 每日：快速回顾昨天的捕获内容（2 分钟）
- 每周：全面回顾 + 连接到现有知识（15 分钟）
- 每月：对旧材料进行留存检查

步骤 2：创建捕获目录

创建 learning/captures/ 用于存放单独的学习笔记。每个来源一个文件：

# [来源标题]
**类型：** 书籍 / 文章 / 课程 / 播客 / 视频 / 经验
**作者/创作者：** [姓名]
**捕获日期：** YYYY-MM-DD
**状态：** 阅读中 / 已完成 / 已放弃
**评分：** [1-5] 价值程度

## 关键要点
1. [最重要的见解]
2. [第二重要的见解]
3. [第三重要的见解]

## 难忘的引述
> "[引述]" — [上下文/页码]

## 这如何应用于我的工作
- [对您的项目/业务的具体应用]
- [与您已知内容的联系]

## 它引发的问题
- [您想进一步探索的事情]

## 回顾历史
| 日期 | 回忆分数 | 备注 |
|------|-------------|-------|
| YYYY-MM-DD | 4/5 | 记住了核心论点，细节模糊 |

步骤 3：添加到 AGENTS.md

## 学习
- 当我分享文章、书籍笔记或学习内容时，将其捕获到 `learning/captures/`
- 在每周回顾期间，提示我回顾到期的项目
- 在相关时，将新学习内容连接到现有捕获
- 在 `learning/learning-state.md` 中追踪我的学习目标
- 当我超过 30 天未重新访问高价值捕获时，建议回顾

使用模式

"我刚刚读了/看了 [文章/书籍/视频]"

智能体创建一个捕获文件，提出有针对性的问题以提取关键要点：

"最重要的单一观点是什么？"
"这与您正在做的工作有何关联？"
"基于此，您会有什么不同的做法？"

"将 [书籍/课程] 添加到我的阅读列表"

智能体使用新条目更新 learning-state.md，并根据当前目标和活跃轨道建议优先级。

"我应该回顾什么？"

智能体检查所有捕获的回顾日期并显示：

从未回顾过的项目（捕获但从未重新访问）
到期间隔重复回顾的项目（7 天 → 30 天 → 90 天）
超过 30 天未回顾的高价值捕获（评分 4-5）

"将此与我已知的内容联系起来"

智能体在现有捕获中搜索相关概念并显示连接：

## 发现的连接
- **[新捕获]** 与 **[现有捕获]** 相关
  - 共享概念：[它们的共同点]
  - 矛盾点：[它们的分歧或提供不同视角之处]
  - 综合见解：[将两者结合得到的收获]

"学习回顾"

每周格式：

## 每周学习回顾 — [日期]

### 本周捕获
| 来源 | 类型 | 评分 | 关键见解 |
|--------|------|--------|-------------|
| [标题] | 文章 | 4/5 | [一句话总结] |

### 回顾提示（间隔重复）
1. **[7 天前的捕获]：** 您能回忆起主要论点吗？[是/否 → 更新回忆分数]
2. **[30 天前的捕获]：** 关键要点是什么？[是/否]
3. **[90 天前的捕获]：** 这如何影响了您的思考？[反思]

### 学习轨道进度
| 轨道 | 上周 | 本周 | 备注 |
|-------|-----------|-----------|-------|
| [主题] | 35% | 42% | 阅读了第 8-10 章 |

### 本周应用
- 将 [来自 X 的概念] 应用于 [项目/决策]
- [来自 Y 的概念] 与我关于 [Z] 的假设相矛盾

### 下周重点
- 优先阅读：[需要关注的内容]
- 到期回顾：[N 个项目需要重新访问]

"就 [主题/捕获] 对我进行测验"

智能体根据捕获内容生成问题以测试留存：

回忆性问题（"X 的 3 种类型是什么？"）
应用性问题（"您如何将 X 应用到您当前的项目中？"）
综合性问题（"X 与您上个月学习的 Y 有何关联？"）

"我学到了关于 [主题] 的什么？"

智能体搜索所有关于某个主题的捕获并进行综合：

## 知识摘要：[主题]

### 来源（按时间顺序）
1. [来源 1] — [日期] — 关键点：[...]
2. [来源 2] — [日期] — 关键点：[...]

### 综合理解
[您对该主题的了解，综合所有来源]

### 思维演变
- 最初认为：[较早的捕获]
- 现在认为：[较晚的捕获，更新的观点]

### 空白点
- 仍不清楚：[遗留的问题]
- 应阅读：[建议的后续来源]

间隔重复计划

该技能使用简单的间隔重复系统：

回顾	之后	目的
第一次	1 天	立即回忆检查
第二次	7 天	短期留存
第三次	30 天	中期留存
第四次	90 天	长期留存
持续	180 天	如果仍然相关则刷新

每次回顾更新回忆分数（1-5）：

5：完美回忆，可以向他人解释
4：良好回忆，细节稍有模糊
3：部分回忆，需要提示
2：模糊回忆，无法解释
1：无回忆，基本上像重新学习

得分为 1-2 的项目将被移到更短的回顾周期。持续得分为 5 的项目可以标记为“已内化”并从活跃回顾中移除。

阅读列表管理

## 阅读列表

### 正在阅读
| # | 标题 | 作者 | 类型 | 开始日期 | 进度 | 优先级 |
|---|-------|--------|------|---------|----------|----------|
| 1 | [书籍] | [作者] | 书籍 | 3/1 | 45% | 🔴 高 |
| 2 | [课程] | [平台] | 课程 | 2/15 | 80% | 🟡 中 |

### 接下来（已排序）
| # | 标题 | 作者 | 类型 | 原因 | 添加日期 |
|---|-------|--------|------|-----|-------|
| 1 | [书籍] | [作者] | 书籍 | [原因/推荐] | 2/28 |
| 2 | [课程] | [平台] | 课程 | [原因] | 3/1 |

### 已完成（最近 90 天）
| 标题 | 类型 | 评分 | 捕获？ | 关键见解 |
|-------|------|--------|----------|-------------|
| [书籍] | 书籍 | 4/5 | ✅ | [一句话总结] |

### 已放弃
| 标题 | 类型 | 放弃原因 | 值得重访？ |
|-------|------|--------------|-------------------|
| [书籍] | 书籍 | 对于当前水平太基础 | 否 |

自定义

学习风格

根据您的学习方式调整捕获格式：

视觉学习者： 包含图表描述、空间关系
实践学习者： 强调“如何应用”部分，包含练习
概念学习者： 专注于框架、心智模型、原则
社交学习者： 记录推荐人、讨论点、教学机会

领域特定模板

为常见来源类型创建捕获模板：

技术书籍： 代码示例部分、实现笔记
商业书籍： 框架摘要、使用时机指南
研究论文： 方法论笔记、局限性、复制潜力
播客： 带时间戳的重点内容、嘉宾背景

知识领域

追踪跨学习领域的广度：

## 知识地图
| 领域 | 捕获数 | 最后活跃 | 深度 |
|--------|----------|-------------|-------|
| 商业策略 | 12 | 2 天前 | 深入 |
| 机器学习 | 3 | 45 天前 | 表面 |
| 写作 | 7 | 10 天前 | 中等 |
| 金融 | 15 | 1 天前 | 深入 |

与其他技能的集成

技能	集成
goal-tracker	将学习目标作为季度 OKR；将捕获进度作为 KR 更新
research-assistant	研究发现自动捕获为学习条目
content-calendar	将高价值捕获转化为内容（教授所学）
writing-assistant	使用捕获作为文章和帖子的素材
decision-log	在决策时参考相关学习内容
habit-tracker	将每日阅读/学习作为习惯追踪

Heartbeat 集成

添加到 HEARTBEAT.md：

## 学习检查
- 是否有任何捕获到期进行间隔重复回顾？
- 距离上次捕获是否超过 7 天？（消费而不捕获 = 浪费）
- 是否有任何学习轨道停滞超过 14 天？

故障排除

问题	解决方法
捕获内容过长	最多限制为 3 个关键要点；深度内容放在参考文献中
从不回顾	设置每周回顾提示的 cron；开始时只处理 3 个项目
活跃轨道过多	同时进行上限为 3 个；在开始新的之前完成或暂停
捕获感觉无用	专注于“如何应用”部分 — 如果无法填写，该来源可能没有价值
阅读列表无限增长	每月清理：移除添加超过 60 天且尚未开始的项目
什么都记不住	前 30 天正常；间隔重复需要 2-3 个周期才能生效

隐私与安全

所有学习数据保留在您的本地工作区
无用于追踪的外部 API 调用（纯基于文件）
捕获可能包含受版权保护的引述 — 仅限个人使用
智能体从不自动分享您的阅读列表或学习笔记

由 The Agent Ledger 开发的 Agent Skills Collection 的一部分。 在 theagentledger.com 订阅以获取高级指南和新技能。

每周安装次数

–

代码仓库

theagentledger/…t-skills

首次出现

–

安全审计

Gen Agent Trust HubPass SocketPass SnykPass

🇺🇸English

Learning Journal

ByThe Agent Ledger — AI-native learning system for people who consume more than they retain.

Your agent tracks what you're learning, prompts you to review key concepts, and makes sure knowledge actually sticks — and gets applied.

The Problem

You read 30 articles a week. You listen to podcasts. You take courses. But 3 months later, you can't recall the key insight from that book that "changed everything." The consumption-to-retention ratio for most people is terrible.

This skill turns your AI agent into a learning partner that captures, organizes, reviews, and connects knowledge over time.

Setup

Step 1: Create State File

Create learning/learning-state.md in your workspace:

# Learning State

## Active Learning Tracks
| Track | Type | Source | Started | Progress | Priority |
|-------|------|--------|---------|----------|----------|
| [Topic] | Book | [Title by Author] | YYYY-MM-DD | 35% | High |
| [Topic] | Course | [Course name — platform] | YYYY-MM-DD | 60% | Medium |
| [Topic] | Self-directed | [Description] | YYYY-MM-DD | Ongoing | Low |

## Learning Goals (This Quarter)
1. [Finish X] — deadline: [date]
2. [Understand Y well enough to Z] — no deadline, but priority
3. [Build skill in Z] — measurable: [what would prove it]

## Review Schedule
- Daily: Quick review of yesterday's captures (2 min)
- Weekly: Full review + connect to existing knowledge (15 min)
- Monthly: Retention check on older material

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

联系我们

Step 2: Create Captures Directory

Create learning/captures/ for individual learning notes. One file per source:

# [Source Title]
**Type:** Book / Article / Course / Podcast / Video / Experience
**Author/Creator:** [name]
**Date captured:** YYYY-MM-DD
**Status:** Reading / Completed / Abandoned
**Rating:** [1-5] how valuable

## Key Takeaways
1. [Most important insight]
2. [Second most important]
3. [Third]

## Memorable Quotes
> "[Quote]" — [context/page]

## How This Applies to My Work
- [Specific application to your projects/business]
- [Connection to something you already know]

## Questions It Raised
- [Things you want to explore further]

## Review History
| Date | Recall Score | Notes |
|------|-------------|-------|
| YYYY-MM-DD | 4/5 | Remembered core thesis, fuzzy on details |

Step 3: Add to AGENTS.md

## Learning
- When I share an article, book note, or learning, capture it in `learning/captures/`
- During weekly reviews, prompt me on items due for review
- Connect new learnings to existing captures when relevant
- Track my learning goals in `learning/learning-state.md`
- Suggest review when I haven't revisited a high-value capture in 30+ days

"I just read [article/book/watched video]"

Agent creates a capture file, asks targeted questions to extract key takeaways:

"What was the single most important idea?"
"How does this relate to what you're working on?"
"What would you do differently based on this?"

"Add [book/course] to my reading list"

Agent updates learning-state.md with new entry, suggests priority based on current goals and active tracks.

"What should I review?"

Agent checks review dates across captures and surfaces:

Items never reviewed (captured but never revisited)
Items due for spaced repetition (7 days → 30 days → 90 days)
High-value captures (rating 4-5) that haven't been reviewed in 30+ days

"Connect this to what I already know"

Agent searches existing captures for related concepts and shows connections:

## Connections Found
- **[New capture]** relates to **[Existing capture]**
  - Shared concept: [what they have in common]
  - Tension: [where they disagree or offer different perspectives]
  - Combined insight: [what you get from both together]

## Weekly Learning Review — [Date]

### This Week's Captures
| Source | Type | Rating | Key Insight |
|--------|------|--------|-------------|
| [Title] | Article | 4/5 | [One-liner] |

### Review Prompts (Spaced Repetition)
1. **[Capture from 7 days ago]:** Can you recall the main argument? [Y/N → update recall score]
2. **[Capture from 30 days ago]:** What was the key takeaway? [Y/N]
3. **[Capture from 90 days ago]:** How has this influenced your thinking? [Reflect]

### Learning Track Progress
| Track | Last Week | This Week | Notes |
|-------|-----------|-----------|-------|
| [Topic] | 35% | 42% | Read chapters 8-10 |

### Applications This Week
- Applied [concept from X] to [project/decision]
- [Concept from Y] contradicted my assumption about [Z]

### Next Week Focus
- Priority reading: [what to focus on]
- Review due: [N items need revisiting]

"Quiz me on [topic/capture]"

Agent generates questions from a capture to test retention:

Recall questions ("What were the 3 types of X?")
Application questions ("How would you apply X to your current project?")
Synthesis questions ("How does X relate to Y that you learned last month?")

"What have I learned about [topic]?"

Agent searches all captures for a topic and synthesizes:

## Knowledge Summary: [Topic]

### Sources (chronological)
1. [Source 1] — [date] — Key point: [...]
2. [Source 2] — [date] — Key point: [...]

### Synthesized Understanding
[What you know about this topic, combining all sources]

### Evolution of Thinking
- Initially thought: [earlier captures]
- Now believe: [later captures, updated views]

### Gaps
- Still unclear on: [questions that remain]
- Should read: [suggested next sources]

Spaced Repetition Schedule

The skill uses a simple spaced repetition system:

Review	After	Purpose
First	1 day	Immediate recall check
Second	7 days	Short-term retention
Third	30 days	Medium-term retention
Fourth	90 days	Long-term retention
Ongoing	180 days	Refresh if still relevant

Each review updates the recall score (1-5):

5: Perfect recall, can explain to others
4: Good recall, minor details fuzzy
3: Partial recall, needed prompting
2: Vague recall, couldn't explain it
1: No recall, basically learning it fresh

Items scoring 1-2 get moved to a shorter review cycle. Items scoring 5 consistently can be marked "internalized" and removed from active review.

Reading List Management

## Reading List

### Currently Reading
| # | Title | Author | Type | Started | Progress | Priority |
|---|-------|--------|------|---------|----------|----------|
| 1 | [Book] | [Author] | Book | 3/1 | 45% | 🔴 High |
| 2 | [Course] | [Platform] | Course | 2/15 | 80% | 🟡 Medium |

### Up Next (Prioritized)
| # | Title | Author | Type | Why | Added |
|---|-------|--------|------|-----|-------|
| 1 | [Book] | [Author] | Book | [Reason/recommendation] | 2/28 |
| 2 | [Course] | [Platform] | Course | [Reason] | 3/1 |

### Completed (Last 90 Days)
| Title | Type | Rating | Capture? | Key Insight |
|-------|------|--------|----------|-------------|
| [Book] | Book | 4/5 | ✅ | [One-liner] |

### Abandoned
| Title | Type | Why Abandoned | Worth Revisiting? |
|-------|------|--------------|-------------------|
| [Book] | Book | Too basic for current level | No |

Adapt capture format to how you learn:

Visual learner: Include diagram descriptions, spatial relationships
Practical learner: Emphasize "How to apply" section, include exercises
Conceptual learner: Focus on frameworks, mental models, principles
Social learner: Note who recommended it, discussion points, teaching opportunities

Domain-Specific Templates

Create capture templates for common source types:

Technical books: Code examples section, implementation notes
Business books: Framework summary, when-to-use guide
Research papers: Methodology notes, limitations, replication potential
Podcasts: Timestamped highlights, guest background

Track breadth across learning domains:

## Knowledge Map
| Domain | Captures | Last Active | Depth |
|--------|----------|-------------|-------|
| Business strategy | 12 | 2 days ago | Deep |
| Machine learning | 3 | 45 days ago | Surface |
| Writing | 7 | 10 days ago | Moderate |
| Finance | 15 | 1 day ago | Deep |

Integration with Other Skills

Skill	Integration
goal-tracker	Learning goals as quarterly OKRs; capture progress as KR updates
research-assistant	Research findings auto-captured as learning entries
content-calendar	Turn high-value captures into content (teach what you learn)
writing-assistant	Use captures as source material for articles and posts
decision-log	Reference relevant learnings when making decisions
habit-tracker	Track daily reading/learning as a habit

Heartbeat Integration

Add to HEARTBEAT.md:

## Learning Check
- Any captures due for spaced repetition review?
- Has it been >7 days since last capture? (consumption without capture = waste)
- Any learning tracks stalled for >14 days?

Issue	Fix
Captures too long	Limit to 3 key takeaways max; depth goes in references
Never reviewing	Set up cron for weekly review prompt; start with just 3 items
Too many active tracks	Cap at 3 simultaneous; finish or pause before starting new ones
Captures feel useless	Focus "How This Applies" section — if you can't fill it, the source may not be valuable
Reading list growing forever	Monthly prune: remove anything added 60+ days ago that you haven't started
Can't recall anything	Normal for first 30 days; spaced repetition needs 2-3 cycles to work

All learning data stays in your local workspace
No external API calls for tracking (pure file-based)
Captures may contain copyrighted quotes — keep for personal use only
Agent never auto-shares your reading list or learning notes

Part of theAgent Skills Collection by The Agent Ledger. Subscribe attheagentledger.com for the premium guide and new skills.

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

83,800 周安装