视频课堂笔记

    1

    Turn Bilibili/YouTube lecture videos into student-style Markdown class notes with timestamps, screenshots, and self-test questions.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    视频课堂笔记

    Example session with this skill installed

    • Read your context and instructions
    • Compiled the output

    我已完成视频《深度学习基础:反向传播算法》的课堂笔记整理。已拉取全长 15:20 的转写稿,并截取了 6 张关键板书图片。笔记文件及截图已保存在 /workspace/deep-learning-notes/ 目录下。

    视频课堂笔记:深度学习基础

    元信息

    • 标题: 深度学习基础:反向传播算法
    • UP/频道: 科技学堂
    • 时长: 15:20
    • 链接: https://www.bilibili.com/video/BV1xP411w7Lz
    • 本课一句话: 深入浅出讲解导数链式法则在神经网络权重更新中的核心应用。

    章节总览

    #时间章节
    100:00–02:15神经网络回顾与损失函数
    202:16–08:45链式法则与梯度计算
    308:46–13:30权重更新与学习率
    413:31–15:20总结与常见误区

    正文内容

    1. 损失函数的定义 00:00–02:15
    • 讲解了 MSE(均方误差)在衡量预测值与真实值差异中的作用。
    • 重要点: 损失函数是一个高维曲面,我们的目标是找到最低点。

    截图 01:45|损失函数公式与梯度下降示意图

    2. 链式法则的应用 02:16–08:45
    • 核心逻辑:复合函数求导。通过局部导数的乘积计算全局梯度。
    • 提到金句:「每一层只负责把误差信号按比例传递给上一层。」

    截图 05:30|链式法则推导过程板书

    自测题

    1. 为什么说反向传播是「链式法则」的直接应用?
    2. 在计算梯度时,如果某一层的激活函数导数为0,会发生什么现象?
    3. 简述学习率过大对收敛过程的影响。

    参考答案

    1. 因为神经网络是嵌套函数,求输出对输入的导数必须逐层相乘。
    2. 梯度消失,导致前层权重无法更新。
    3. 会导致损失函数在最小值附近震荡甚至发散。

    Next steps

    • 将生成的 .md 文件导入您的 Obsidian 库。
    • 查看 /workspace/deep-learning-notes/ 下的高清板书截图。
    • 尝试完成剩余 7 道深度自测题以巩固理解。

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Stop manually summarizing video content and start building a structured knowledge base. This skill acts as a dedicated student, transforming Bilibili and YouTube links into comprehensive, formatted study notes. It handles the heavy lifting of timestamping, multi-part video organization, and visual documentation so you can focus on learning.

    What it does

    • Structured Note Taking creates Markdown files with metadata, chapter tables, and detailed body text using a student perspective.
    • Visual Documentation captures screenshots of key slides, definitions, and quotes, embedding them directly into the notes.
    • Timestamp Mapping tracks every knowledge point with precise time ranges, supporting both single videos and multi-part (P1/P2) series.
    • Active Recall Generation produces 8-12 self-test questions ranging from basic concepts to application, with a separate answer key at the end.
    • Multi-Platform Support processes Bilibili, YouTube, short links, and local video files including metadata like UP/Channel and duration.

    How it works

    1. Link Parsing extracts metadata, expands short URLs, and maps out multi-part video structures.
    2. Context Retrieval pulls ASR transcripts or subtitles and aligns them with video chapters for time-code accuracy.
    3. Evidence Capture identifies key frames to generate screenshots named by timestamp and topic.
    4. Markdown Synthesis compiles the final .md file with Obsidian-compatible image paths and structured self-test sections.

    Frameworks & tools

    Uses FFmpeg for frame extraction and supports Markdown editors like Obsidian or log-based systems like WPS and Feishu Wiki.

    Why this beats prompting it yourself

    Generic prompts often miss specific timestamps or fail to capture visual evidence from the video itself. This skill enforces a consistent schema including metadata headers, linked screenshots, and separated self-test sections that are ready for immediate archival.

    Use cases

    • Online Course Archiving to turn a 20-part Bilibili series into a searchable Markdown library.
    • Technical Tutorial Review by capturing code snippets and diagrams directly from the screen with specific time references.
    • Meeting/Webinar Summarization to extract key arguments and visual slides for team distribution.

    Known limitations

    Requires access to transcripts or ASR for precise timestamping. Screenshot quality depends on source video resolution.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.

    Requires transcript/ASR access for time-coding. Output assumes Markdown-compatible environment like Obsidian for image rendering. Link expansion requires active web access.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 10 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed10 days ago

    What's inside

    Frequently Asked Questions

    The skill creates structured Markdown notes including metadata (UP/Channel, duration), timestamped chapter summaries, embedded screenshots of key slides, and a dedicated self-test section for active recall.

    Yes, the skill is designed to parse multi-part video series (e.g., P1, P2) and long-form content, maintaining consistent organization across the entire playlist or video set.

    While optimized for Markdown editors like Obsidian, the formatted text and images are fully compatible with any tool that supports Markdown, including Notion, Feishu Wiki, and WPS.

    The skill relies on available transcripts or ASR (Automatic Speech Recognition) to map timestamps; if a video lacks these, the accuracy of the chronological summary may be limited.

    Unlike generic prompts, this skill uses FFmpeg for precise visual evidence capture and enforces a consistent schema that ensures screenshots are correctly named and linked within your notes.

    Your purchase includes the full skill configuration ready for deployment on supported AI agent platforms, along with any necessary documentation for linking your video processing environments.

    New in Agents & Orchestration

    Free