ytdl

Convert a video/podcast URL into a searchable markdown entry in the llm-wiki.

Steps:

1. If $ARGUMENTS is empty, ask: "What URL? (YouTube, podcast, any yt-dlp-supported source)".

2. Verify prerequisites with Bash:

   • command -v yt-dlp — if missing, print pipx install yt-dlp and stop.
   • command -v whisper-cli or command -v whisper.cpp or check ~/whisper.cpp/main — if missing, note transcription is skipped (still downloads audio).

3. Invoke the helper script which already handles the pipeline:

   bash ~/Desktop/ACTIVITIES/claude-universal/scripts/ytdl-to-wiki.sh "$ARGUMENTS"
   

   If the helper is missing (first-time machine), fall back to:

   slug=$(echo "$ARGUMENTS" | sha1sum | cut -c1-8)
   out="$HOME/Desktop/ACTIVITIES/llm-wiki/media/$(date +%Y%m%d)-$slug"
   mkdir -p "$out"
   yt-dlp --write-info-json --write-description -x --audio-format mp3 \
     -o "$out/%(title)s.%(ext)s" "$ARGUMENTS"
   # transcribe first .mp3
   whisper-cli -f "$out"/*.mp3 -otxt -of "$out/transcript" 2>/dev/null || true
   

4. Produce the markdown entry at ~/Desktop/ACTIVITIES/llm-wiki/media/<date>-<slug>.md:

   # <video title>
   
   **Source:** $ARGUMENTS
   **Downloaded:** YYYY-MM-DD
   **Duration:** <from info.json>
   **Uploader:** <from info.json>
   
   ## Description
   <from yt-dlp --write-description>
   
   ## Transcript
   <whisper output, or "Transcription skipped — whisper.cpp not installed">
   
   ## Key takeaways
   <if transcript exists, give 5 bullet summary>
   

5. Append a pointer to today's wiki file: - 🎥 [<title>](media/<date>-<slug>.md) — <url>

6. Print: ✓ ytdl complete — llm-wiki/media/<date>-<slug>.md.

Large videos (>30 min): warn the user before transcribing and ask to proceed.