agentsclimarketplace
Sign in

Yt to md

Skill felipefrf/yt-to-md/skills/yt-to-md

Convert YouTube videos into structured Markdown transcripts for AI agent consumption. Use when the user provides a YouTube URL and wants to extract the video content, transcript, chapters, or metadata as Markdown. Also use when the user wants to feed YouTube video content into other agents or workflows.From its SKILL.md

Install
npx -y skills add felipefrf/yt-to-md --skill yt-to-md

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.8 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

What I do

I convert YouTube videos into well-structured Markdown documents containing:

  • Video metadata (title, uploader, date, duration, views, tags)
  • Full transcript with optional timestamps
  • Chapter markers (when available)

When the user asks for a summary, you (the agent) produce the summary yourself — read the generated markdown file and write a summary section at the end of it. There is no external API key needed; you are the LLM.

When to use me

  • User provides a YouTube URL and asks to "convert", "extract", "transcribe" or "summarize" it
  • User wants to feed a video's content into another agent, analysis tool, or LLM workflow
  • User asks to download captions/transcripts from YouTube
  • User mentions yt-to-md or requests YouTube-to-markdown conversion

CLI reference

yt-to-md <url> [--lang pt] [--output-dir ./dir] [--no-timestamps] [--filename NAME]
yt-to-md --batch urls.txt [--delay 3] [--output-dir ./dir]
FlagDescription
urlYouTube URL or video ID (required)
--lang LANGPreferred subtitle language (default: tries pt, en, es)
--output-dir DIROutput directory (default: current dir)
--no-timestampsOmit timestamps from transcript
--filename NAMECustom output filename (without .md extension). Ignored in batch mode.
--batch FILEProcess a file of URLs (one per line, # comments ok) in series. Use for volume to avoid IP rate-limiting.
--delay SECONDSPause between videos in batch mode (default: 3.0). Raise if you hit rate limits.

How it works (fallback chain)

StepMethodNotes
Transcriptyoutube-transcript-api (primary)Lightweight, no blocking. Falls back to yt-dlp.
Metadatayt-dlp --dump-json (primary)Falls back to YouTube oEmbed API (title, author, thumbnail).
Chaptersyt-dlp --dump-jsonOnly available when yt-dlp is not blocked.

Workflow

Step 1: Fetch the video

Run the CLI tool. It prints the output path to stdout and progress to stderr.

yt-to-md <url> [--lang LANG] [--output-dir DIR]

Dependencies: The tool auto-detects available backends:

  • youtube-transcript-api (essential for transcripts)
  • yt-dlp (optional, enriches metadata and chapters)

If yt-dlp is blocked (YouTube anti-bot), the tool falls back to the oEmbed API for basic metadata — transcripts still work via youtube-transcript-api.

Workflow

Step 1: Fetch the video

Run the CLI tool. It prints the output path to stdout and progress to stderr.

yt-to-md <url> [--lang LANG] [--output-dir DIR]

If yt-dlp is not installed, tell the user to run:

pip install yt-dlp

If there are no subtitles available, the tool still produces a valid markdown file with metadata and description — proceed with what you have.

Step 2: Read the generated file

The tool prints the path to the generated .md file on stdout. Read it immediately.

Step 3: Summarize (if the user asked for a summary)

You are the LLM — generate the summary yourself. Do NOT use --summarize or call an external API.

  1. Read the full generated markdown file
  2. Write a summary with these sections:
    • Overview (2-3 sentences)
    • Key Points (bullet list)
    • In-depth Summary (if transcript is available, paragraph-by-paragraph)
    • Takeaways (actionable insights)
  3. Append the summary at the end of the file, under an ## AI Summary heading, followed by a separator --- and *Summary generated by <your model name>* — use your own model identity (e.g. Opus 4.8, Sonnet 5, DeepSeek V4 Pro), not the tool or harness name.

Step 4: Present to the user

Show a concise overview of what was extracted:

  • Video title, uploader, duration
  • Number of transcript entries
  • Whether chapters were found
  • Output file path
  • If summarization was done, note it

Example output structure

# Video Title

> **Source:** YouTube Video Transcript
> **Tool:** yt-to-md v0.1.0
> **URL:** https://...
> **Video ID:** dQw4w9WgXcQ

---

## Video Metadata

| Field | Value |
|-------|-------|
| **Title** | ... |
| **Uploader** | ... |
| **Date** | ... |
| **Duration** | ... |

---

## Description

Full video description from YouTube...

---

## Chapters

- `00:00` Introduction
- `02:30` Main Topic

---

## Transcript

### 00:00 — Introduction

Full transcript text here...

---

## AI Summary

> *Summary generated by Opus 4.8*

### Overview
...

### Key Points
- ...

### Takeaways
- ...

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most video audio skills give in ~1.2k tokens

Counted across 622 of the 795 authors here whose files we hold, read 2026-08-07

  • Read individual rule files for detailed explanationsin 21 of 622, across 10 files
  • Render final videoin 13 of 622, across 6 files
  • Use WAV PCM 16kHz mono audio formatin 12 of 622, across 3 files
  • Use this skill when dealing with Remotion codein 11 of 622, across 4 files
  • Save generated audio to a WAV filein 11 of 622, across 4 files
  • Handle conversion errors gracefullyin 10 of 622, across 6 files
  • Add captions to videos alwaysin 10 of 622, across 4 files
  • Generate music from text descriptions using MusicGenin 9 of 622, across 2 files
  • Do not skip pipeline layersin 9 of 622, across 3 files
  • Do not make one tool do everythingin 9 of 622, across 3 files
  • Use Azure Document Intelligence for complex PDFsin 9 of 622, across 4 files
  • Never ask the user to paste their full API keyin 9 of 622, across 3 files

Said here and by no other author read

  • Run yt-to-md to fetch the video content
  • Read the generated markdown file immediately
  • Generate summaries yourself using your own reasoning
  • Append AI summaries under an AI Summary heading
  • Use your own model identity in summary attributions
  • Increase the delay if you hit rate limits

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.