wx-video-account-notes

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

微信视频号下载与笔记整理

WeChat Channels Video Download and Note Organization

处理一个视频号分享链接，生成可供 agent 继续整理的结构化笔记材料。

Process a WeChat Channels share link to generate structured note materials for further organization by agents.

触发条件

Trigger Conditions

仅在用户提供

https://weixin.qq.com/sph/...

视频号分享链接，并要求继续处理时使用。

This tool is only used when the user provides a WeChat Channels share link in the format

https://weixin.qq.com/sph/...

and requests further processing.

输入

Inputs

必需：一个视频号分享链接
可选：输出目录
可选：期望标题或文件名

Required: A WeChat Channels share link
Optional: Output directory
Optional: Desired title or filename

固定流程

Fixed Workflow

Windows 运行

pwsh -File "scripts/bootstrap.ps1"

；macOS/Linux 运行

sh scripts/bootstrap.sh

。

Windows 运行

pwsh -File "scripts/invoke_pipeline.ps1" -ShareUrl "https://weixin.qq.com/sph/..."

；macOS/Linux 运行

sh scripts/invoke_pipeline.sh --share-url "https://weixin.qq.com/sph/..."

。

pipeline 负责下载视频或图文动态图片、抽帧、OCR、ASR，生成
```
note_materials.json
```
。
当前 agent / model 基于
```
note_materials.json
```
写最终
```
<slug>.md
```
。
如果用户明确说不要笔记，跳过最终
```
<slug>.md
```
。

On Windows, run

pwsh -File "scripts/bootstrap.ps1"

; on macOS/Linux, run

sh scripts/bootstrap.sh

On Windows, run

pwsh -File "scripts/invoke_pipeline.ps1" -ShareUrl "https://weixin.qq.com/sph/..."

; on macOS/Linux, run

sh scripts/invoke_pipeline.sh --share-url "https://weixin.qq.com/sph/..."

The pipeline is responsible for downloading videos or images from graphic posts, frame extraction, OCR, ASR, and generating
```
note_materials.json
```
.
The current agent/model writes the final
```
<slug>.md
```
based on
```
note_materials.json
```
.
Skip generating the final
```
<slug>.md
```
if the user explicitly states they don't want the note.

输出约定

Output Convention

text

<output-dir>/
  <slug>.mp4        # 仅视频动态
  note_materials.json
  raw.json
  ocr.txt
  asr.txt
  ocr_frames/
  frames/
  audio/

默认情况下，当前 agent / model 还应写出同目录的

<slug>.md

。如果用户明确说不要笔记，跳过该文件。

text

<output-dir>/
  <slug>.mp4        # Video posts only
  note_materials.json
  raw.json
  ocr.txt
  asr.txt
  ocr_frames/
  frames/
  audio/

By default, the current agent/model should also write

<slug>.md

in the same directory. Skip this file if the user explicitly says they don't want the note.

成稿规则

Drafting Rules

以
```
note_materials.json
```
为主。
必要时对照
```
raw.json
```
、
```
ocr.txt
```
、
```
asr.txt
```
。
参考
```
resources/model-note-template.md
```
。
不要编造提取结果中不存在的信息。
如果
```
note_materials.json
```
包含
```
visual_frames
```
字段，且当前模型有识图能力，必须读取
```
visual_frames
```
对应图片辅助理解视频内容。
对图文动态，
```
visual_frames
```
是原始图片素材；如果当前模型有识图能力，必须读取这些图片补充 OCR 无法表达的版式、界面、图表或视觉上下文。

Prioritize using
```
note_materials.json
```
.
Cross-reference with
```
raw.json
```
,
```
ocr.txt
```
,
```
asr.txt
```
when necessary.
Refer to
```
resources/model-note-template.md
```
.
Do not fabricate information that does not exist in the extracted results.
If
```
note_materials.json
```
contains the
```
visual_frames
```
field and the current model has image recognition capabilities, you must read the corresponding images in
```
visual_frames
```
to assist in understanding the video content.
For graphic posts,
```
visual_frames
```
are the original image materials; if the current model has image recognition capabilities, you must read these images to supplement layout, interface, charts, or visual context that cannot be expressed by OCR.

失败处理

Failure Handling

如果自举失败，停止并明确说明失败的资产或命令。
如果 OCR 或 ASR 失败，在材料和最终 Markdown 中明确标记失败。
不要把失败结果伪装成成功。

If bootstrap fails, stop and clearly state the failed asset or command.
If OCR or ASR fails, clearly mark the failure in the materials and final Markdown.
Do not disguise failure results as success.

实施要求

Implementation Requirements

优先使用 skill 目录下当前平台的
```
.runtime/<os>-<arch>/
```
环境，不依赖系统 PATH。
低层下载、抽帧、抽音频、OCR、ASR 由脚本和 Python 完成。
视频动态中，
```
ocr_frames/
```
存放 ffmpeg 直接裁出的字幕区域小图；
```
frames/
```
只保留少量完整参考帧，供
```
visual_frames
```
使用。
图文动态中，
```
frames/
```
存放下载到的原始图片，供 OCR 和
```
visual_frames
```
使用；该类动态跳过 ASR。
当前 agent / model 只负责基于材料整理最终
```
<slug>.md
```
。

Prioritize using the
```
.runtime/<os>-<arch>/
```
environment for the current platform in the skill directory, without relying on the system PATH.
Low-level tasks such as downloading, frame extraction, audio extraction, OCR, and ASR are completed by scripts and Python.
For video posts,
```
ocr_frames/
```
stores small cropped images of subtitle areas directly extracted by ffmpeg;
```
frames/
```
only retains a small number of complete reference frames for use in
```
visual_frames
```
.
For graphic posts,
```
frames/
```
stores the downloaded original images for OCR and
```
visual_frames
```
; skip ASR for this type of post.
The current agent/model is only responsible for organizing the final
```
<slug>.md
```
based on the materials.