wx-video-account-notes

Original：🇨🇳 Chinese

Translated

30 scripts

Process WeChat Channels share links, extract OCR/ASR note materials and generate Markdown notes.

4installs

Added on2026-08-12

NPX Install

npx skill4agent add muzijin701/wx-video-account-notes wx-video-account-notes

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

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.

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.

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.

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.

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.

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.