Huo15 Openclaw Mit 48h Learning Method

v2.2.2

麻省理工学院48小时学习法技能(青岛火一五信息科技有限公司)。使用 NotebookLM CLI 实现 MIT 研究生 Ihtesham Ali 的三问学习框架: 1. 问心智模型:领域内专家共享的 5 个基本思维框架 2. 问专家分歧:在哪 3 个问题上根本不同意 3. 问暴露性问题:生成能区分真懂和假背的 1...

1· 170·0 current·0 all-time
byJob Zhao@zhaobod1

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for zhaobod1/huo15-openclaw-mit-48h-learning-method.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Huo15 Openclaw Mit 48h Learning Method" (zhaobod1/huo15-openclaw-mit-48h-learning-method) from ClawHub.
Skill page: https://clawhub.ai/zhaobod1/huo15-openclaw-mit-48h-learning-method
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install huo15-openclaw-mit-48h-learning-method

ClawHub CLI

Package manager switcher

npx clawhub@latest install huo15-openclaw-mit-48h-learning-method
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The skill's name/description (MIT 48‑hour learning method using NotebookLM) match the actual behavior: it calls a local NotebookLM CLI (nlm), creates/listens to notebooks, adds sources, runs prompts, and produces audio/video. The CLI login via Google is expected because NotebookLM requires Google authentication. No unrelated cloud provider credentials or extraneous binaries are requested.
Instruction Scope
SKILL.md and the script instruct the agent/user to run and re-run nlm login, add local files/URLs/YouTube, and save notebook IDs. This is within scope for a NotebookLM automation tool. Note: the script auto-detects login expiry and will re-run 'nlm login' (interactive browser auth) without extra confirmation, and it writes state files (~/.mit-learn-notebook-id and ~/.mit-learn-env) into the user's home directory — behavior you should expect but be aware of.
Install Mechanism
No install spec is provided; the skill assumes an existing NotebookLM CLI at ~/.venv/notebooklm/bin/nlm. No external downloads or archives are fetched by the skill itself, which keeps install risk low. The script does call python3 for URL decoding, which is a reasonable dependency.
Credentials
The registry lists no required secrets. The script uses optional env vars NOTEBOOKLM_PROFILE and MIT_LEARN_LANG (both reasonable). Authentication happens through the nlm login flow (Google account) which is appropriate for NotebookLM integration and not requested directly as raw tokens in the skill.
Persistence & Privilege
always:false (normal). The script writes small state files (~/.mit-learn-notebook-id and ~/.mit-learn-env) and may append a NOTEBOOK_ID line to ~/.mit-learn-env. It does not modify other skills or system-wide configs. The auto-login behavior increases convenience but means the script may trigger interactive login flows when tokens expire.
Assessment
This skill wraps the NotebookLM CLI to implement the MIT "three questions" learning workflow. Before installing or running it: 1) verify you have a trusted NotebookLM CLI at the configured path (default ~/.venv/notebooklm/bin/nlm); 2) expect an interactive Google sign-in (nlm login) and that the script may re-run that login when auth expires; 3) the script will create/update files in your home directory (~/.mit-learn-notebook-id and ~/.mit-learn-env) — inspect these files if you are concerned about local state; 4) review scripts/mit-learn.sh yourself (it is included) to confirm no additional endpoints or unexpected commands are present; and 5) run in a user-level environment (not as root) and ensure python3 and standard CLI tools are from trusted sources. If you need stricter control, run the commands manually rather than granting the agent autonomous invocation.

Like a lobster shell, security has layers — review code before you run it.

latestvk97e35ab356fpscgk8e6s8ntmd85fhmn
170downloads
1stars
3versions
Updated 4d ago
v2.2.2
MIT-0

火一五 MIT 48 小时学习法

MIT 研究生 Ihtesham Ali 的学习方法:48 小时内通过三问框架掌握任意领域。

核心工作流

学什么 → 创建 NotebookLM → 添加资料 → 三问框架 → 生成 Audio/Video

前置条件

首次使用必须认证:

~/.venv/notebooklm/bin/nlm login

(会打开浏览器,按提示完成 Google 账号授权)

自动续登录: 脚本会在每次执行命令前自动检测登录状态,如果检测到登录已失效,会自动重新运行 nlm login,无需手动干预。

依赖

  • CLI 工具~/.venv/notebooklm/bin/nlm
  • 环境变量NOTEBOOKLM_PROFILE(可选,默认为 default
  • 语言设置MIT_LEARN_LANG(可选,默认为 zh-CN

脚本位置

skills/huo15-mit-48h-learning-method/scripts/mit-learn.sh

使用方法

完整流程(推荐)

./scripts/mit-learn.sh full "学习主题" --url "https://..." --file ./notes.pdf --youtube "https://youtube.com/..."

完整流程包含:创建 notebook → 添加资料 → 三问框架(心智模型、专家分歧、暴露性问题)

分步流程

# 1. 创建笔记本
./scripts/mit-learn.sh init "机器学习基础"

# 2. 添加资料(可多次调用)
./scripts/mit-learn.sh add --url "https://..." --wait
./scripts/mit-learn.sh add --file ./paper.pdf --wait
./scripts/mit-learn.sh add --youtube "https://youtube.com/..."

# 3. 三问框架
./scripts/mit-learn.sh ask mental-models     # 问心智模型(5个框架)
./scripts/mit-learn.sh ask disagreements     # 问专家分歧(3个问题)
./scripts/mit-learn.sh ask probing           # 问暴露性问题(10个问题)
./scripts/mit-learn.sh ask all               # 完整三问

# 4. 生成概览
./scripts/mit-learn.sh audio                 # 生成播客音频
./scripts/mit-learn.sh video                 # 生成视频

# 5. 查看状态
./scripts/mit-learn.sh status                # 查看当前 notebook 状态
./scripts/mit-learn.sh list                  # 列出所有 notebooks

三问框架详解

问心智模型(Mental Models)

"该领域专家共享的 5 个基本思维框架是什么?"

  • 每个框架用一句话解释 + 具体应用例子
  • 目的是快速建立领域内专家共同认可的思维工具箱

问专家分歧(Expert Disagreements)

"在哪 3 个问题上,该领域专家根本不同意?"

  • 识别核心理论、方法或结论上的根本性争议
  • 了解分歧根源,明白这不是细枝末节而是根本矛盾
  • 这是区分真学习和假学习的关键:知道分歧意味着真正理解领域

问暴露性问题(Probing Questions)

"生成 10 个能区分真懂和假背的问题"

  • 苏格拉底式追问:开放性问题,无法通过简单回忆回答
  • 每个问题需说明:假背者会怎么错 / 真懂的人会怎么答
  • 这是检验学习效果的最终武器

NotebookLM 支持的资料类型

类型参数示例
URL--url / -u--url "https://..."
文件--file / -f--file ./notes.pdf
YouTube--youtube / -y--youtube "https://youtube.com/..."
Google Drive--drive--drive <doc-id>

Audio/Video 选项

Audio(播客音频)

./scripts/mit-learn.sh audio [format]
# format: deep_dive(默认)/ brief / critique / debate
# length: short / default / long

Video(视频概览)

./scripts/mit-learn.sh video [style]
# style: auto_select(默认)/ classic / whiteboard / kawaii / anime / watercolor / retro_print / heritage / paper_craft

提示词设计原则

三问框架的提示词基于以下原则设计:

  1. 心智模型:要求专家视角 + 具体例子,不可泛泛而谈
  2. 专家分歧:要求根本性分歧,而非表面差异
  3. 暴露性问题:苏格拉底追问法,必须能区分真假理解

注意事项

  • init 后当前 notebook ID 保存到 ~/.mit-learn-notebook-id,后续命令复用
  • 添加资料后可用 --wait 等待处理完成
  • NotebookLM API 可能有速率限制,避免短时间内大量请求
  • 三问结果建议保存到笔记中,用于后续复习
  • 如果使用多个 Google 账号,可设置 NOTEBOOKLM_PROFILE=your-profile 环境变量切换

v2.0.0 (2026-04-06)

新功能

  • 支持 file:// URL:自动转换为真实路径再添加
  • 音频生成等待:新增 wait_for_audio 确认音频生成完成再返回
  • 重复 notebook 检测:创建前先查找同名 notebook,避免重复
  • 自动续登录:登录失效前自动重新运行 nlm login,无需手动干预

Bug 修复

  • full 命令参数传递 bug:修复了 urls/files/yt_urls 三个数组错误传递的问题
  • 增强错误处理:cmd_add 不再隐藏错误信息,现在会明确显示失败原因

改进

  • 新增 --skip-audio flag:full 命令可跳过音频生成
  • cmd_init 重构为 get_or_create_notebook 函数
  • wait_for_processing 稳定性提升
  • debug 函数默认关闭,减少干扰
mit-learn.sh full "机器学习" --file ./notes.pdf --skip-audio

Comments

Loading comments...