YouTube Playlist Curator Skill (Dynamic $N$-Subagent Fan-Out)

When this skill is invoked via /urlPlaylist <url> or /okc-urlPlaylist <url>, execute the following dynamic $N$-subagent workflow:

1. Turn 1 — Concurrent Playlist Extraction & Pre-Flight Discovery (FastMCP):
   • Execute the native FastMCP tool fetch_playlist_data(url="<url>", skip_existing=True) on the okc server (do not spawn CLI commands unless MCP is unavailable).
   • This scrapes the playlist structure via yt-dlp --flat-playlist, checks vault_index.db in <15ms for any video IDs already curated (already_exists), fetches all pending video transcripts concurrently via ThreadPoolExecutor(max_workers=6), maps graphify_context, and returns:
     • total_videos: Total videos in the playlist.
     • already_indexed_count: Videos already present in the vault.
     • pending_count ($N$): Number of new videos requiring curation.
     • recommended_subagents: Number of parallel subagents to launch (0, 1..8).
     • videos: List of video objects with index, id, title, url, duration, already_exists, existing_note, transcript (inline for $\le 5$ pending videos), and transcript_path_abs (temp/playlist_transcripts/XX_VIDEOID.txt).
2. Turn 2 — Dynamic $N$-Subagent Parallel Drafting (1 Video = 1 Subagent):
   • Inspect pending_count ($N$) from the fetch_playlist_data payload:
     • If $N == 0$: All videos in the playlist are already curated in the vault. Report their existing [[wikilinks]] and stop.
     • If $N == 1$: Draft the single pending video note directly in the main agent and commit via commit_curated_note(..., fast_sync=True, async_sync=True).
     • If $2 \le N \le 8$: Spawn $N$ parallel subagents (invoke_subagent) in a single turn — 1 subagent per pending video.
       • Provide each subagent with its video metadata (index, id, title, channel, url, playlist_title), graphify_context (author_existing_notes, matched_wiki_notes), target filename (<Series> ## — <Title>.md), and either its inline transcript or transcript_path_abs path to read.
       • Each subagent must return ONLY the complete formatted raw/ Markdown note (no YAML, H1 title, canonical blockquote with Processed: DD-MM-YYYY and Tags: #no-read-yet, Key Takeaways, numbered sections, Mermaid diagram, Flashcards, Glossary, Related wikilinks) and 2–3 candidate wiki/ concept summaries.
     • If $N > 8$: Dispatch subagents in parallel waves of $6\text{–}8$ concurrent subagents (1 subagent per video per wave) until all $N$ pending videos are drafted.
3. Turn 3 — Consolidate & Atomically Commit Batch (FastMCP):
   • Collect all $N$ drafted raw/ notes from the subagents.
   • Generate a dedicated Series Master Plan (Master Plan — <Series Name>.md) mapping the complete playlist (including both newly curated notes and any pre-existing already_exists notes), conceptual architecture, and table of notes.
   • Synthesize 3–7 core wiki/ concepts across the series.
   • Atomically commit all raw/ notes, the Series Master Plan, and wiki/ concepts, surgically upsert into SQLite, and trigger the detached background post-commit worker (async_post_commit.py) in a single call using FastMCP commit_curated_batch:

     commit_curated_batch(
         raw_notes={
             "dataScienceKnowledgeBase/AI Engineer/raw/<Series>/<Series> 01 — Title.md": "...",
             "dataScienceKnowledgeBase/AI Engineer/raw/<Series>/<Series> 02 — Title.md": "...",
         },
         master_plan_path="dataScienceKnowledgeBase/AI Engineer/raw/<Series>/Master Plan — <Series>.md",
         master_plan_content="...",
         wiki_notes={"ConceptName.md": "..."},
         target_kb="dataScienceKnowledgeBase/AI Engineer",
         fast_sync=True,
         async_sync=True
     )

   • Report progress and display clickable file:// and Obsidian URI links for all created notes.