YouTube Transcript Fetcher

Fetches subtitle text from a YouTube video and outputs it as Markdown. The retrieved transcript can then be used for downstream tasks such as summarization, analysis, translation, or answering questions about the video.

Prerequisites

• uv must be installed on the system (used for dependency resolution and script execution)
• Network access to YouTube

Usage

uv run <skill-dir>/scripts/fetch_transcript.py "<YouTube URL>"

Output goes to stdout in Markdown format. Pipe or redirect as needed:

# Capture to a variable
TRANSCRIPT=$(uv run <skill-dir>/scripts/fetch_transcript.py "https://www.youtube.com/watch?v=XXXXXXXXXXX")

# Save to a file
uv run <skill-dir>/scripts/fetch_transcript.py "https://youtu.be/XXXXXXXXXXX" > transcript.md

Supported URL formats

The script auto-extracts the video ID from any of these:

• https://www.youtube.com/watch?v=XXXXXXXXXXX
• https://youtu.be/XXXXXXXXXXX
• https://www.youtube.com/shorts/XXXXXXXXXXX

Output format

# Video Title

- **Channel**: Channel Name
- **URL**: Original URL

## Transcript (language_code)

Full subtitle text joined with spaces as plain text.

Language priority

Transcript lookup follows this order:

1. English (en) or Japanese (ja) manual captions
2. If neither is found, the first available transcript track

Auto-generated captions are included. Videos with no captions at all will error.

Error cases

The script prints to stderr and exits non-zero on failure. Common causes:

• Cannot extract video ID — unsupported or malformed URL
• Metadata fetch failed — video is private, deleted, or network error
• Transcript fetch failed — captions are disabled or none exist

When an error occurs, inform the user and suggest they check the video's visibility and caption settings.

Typical workflow

When a user shares a YouTube link and asks about its content:

1. Run the script to fetch the transcript
2. Read the output to understand the video content
3. Perform the requested task (summarize, analyze, answer questions, etc.)