agentleFS
TheVellerTheVeller/claude-skills/skills/research/research-pipeline/SKILL.md

research-pipeline

Skill4 starsChanged 14 days agoMITView on GitHub

YouTube search + NotebookLM analysis super-skill — searches YouTube, sends videos to NotebookLM for server-side deep analysis, optionally generates a deliverable (infographic, podcast, slides), saves to the vault. Zero Claude tokens for analysis. Use when researching a topic via YouTube or generating research deliverables from video content.

17 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use experiment-bridge, resubmit-pipeline and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • raja21068Skill

    experiment-bridge

    Workflow 1.5: Bridge between idea discovery and auto review. Reads EXPERIMENT_PLAN.md, implements experiment code, deploys to GPU, collects initial results. Use when user says \"实现实验\", \"implement experiments\", \"bridge\", \"从计划到跑实验\", \"deploy the plan\", or has an experiment plan ready to execute.

    Also in 4 projects that use this

  • raja21068Skill

    resubmit-pipeline

    Workflow 5: orchestrate a text-only resubmit of a polished paper to a different venue under hard constraints (no new experiments, no bib edits, no framework changes, never overwrite prior submissions). Phase 0 physical isolation, Phase 0.5 health + anonymity check, Phase 1 audit (proof / claim / citation), Phase 2 microedits via auto-loop with edit-whitelist + citation-audit --soft-only, Phase 3 kill-argument adversarial gate, Phase 4 final compile + Overleaf push via /overleaf-sync. Use when user says \"resubmit pipeline\", \"重投流程\", \"port paper to <new venue>\", \"resubmit to <venue>\", \"tighten paper for resubmission\", or has a rejected/withdrawn paper to move to a different top venue under tight time budget.

    Also in 4 projects that use this

  • rongaredeSkill

    semantic-scholar

    使用 Semantic Scholar API 检索和验证学术论文。支持并发多关键词搜索、批量 ID/DOI 查询、引用分析、批量补全摘要、统一 Markdown 导出。覆盖 2.14 亿+ 学术论文,无需 API Key 即可使用。触发词:论文检索、论文验证、Semantic Scholar、S2 搜索、查论文、补全摘要、导出MD

    Also in 5 projects that use this

  • +9 more from raja21068

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 7 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user wants to:

    line 16
  2. Distinctive:

    Do content research (what's working, view patterns, gaps to exploit)

    line 20
  3. Distinctive:

    Search YouTube for the query and collect video URLs + metadata.

    line 33
  4. Distinctive:

    Create a dedicated notebook for this research session:

    line 45
  5. Distinctive:

    Add each video URL as a source to the notebook (up to 50):

    line 55
  6. Distinctive:

    Add all videos in sequence.

    line 61
  7. Distinctive:

    If user requested a specific deliverable, run in parallel with or after Step 5:

    line 91
  8. Very specific:

    Use --wait flag to block until deliverable is ready

    line 171

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Research Pipeline Skill

A super-skill that combines YouTube search + NotebookLM analysis into a single research workflow. Searches YouTube for videos, sends them to NotebookLM for deep analysis, optionally generates a deliverable (infographic, podcast, slide deck, etc.), and saves everything to the Obsidian vault.

Token Budget

This pipeline is designed to use zero Claude tokens for analysis. NotebookLM processes the corpus server-side (Google's infrastructure). Claude's only job is to orchestrate the pipeline, capture notebooklm ask stdout, and write the result to Obsidian. Never re-read large scraped documents in Claude — pass them directly to NotebookLM and let it do the analysis.

Trigger

Use this skill when the user wants to:

  • Research a topic using YouTube as the source
  • Analyze trends, gaps, and insights from YouTube videos on a subject
  • Generate research deliverables (infographic, podcast, summary) from YouTube content
  • Do content research (what's working, view patterns, gaps to exploit)

Explicit triggers:

  • /research-pipeline
  • "research [topic] on YouTube"
  • "analyze YouTube videos about [topic]"
  • "YouTube pipeline for [topic]"
  • "find and analyze top videos about [topic]"

Full Pipeline

Step 1 — YouTube Search (youtube-search skill)

Search YouTube for the query and collect video URLs + metadata.

yt-dlp "ytsearch{N}:{QUERY}" --dump-json --no-download --flat-playlist --no-warnings 2>/dev/null

Parse results → extract video IDs → build URLs.

Default: 10 videos. User can specify (e.g., "top 5", "top 20").

Step 2 — Create NotebookLM Notebook

Create a dedicated notebook for this research session:

notebooklm create "{QUERY} Research - {DATE}"

Save the notebook ID from output.

Step 3 — Add YouTube Sources

Add each video URL as a source to the notebook (up to 50):

notebooklm source add "https://www.youtube.com/watch?v={ID}"

Add all videos in sequence. NotebookLM will process the transcripts automatically.

Step 4 — Wait for Processing

Sources take 30-60 seconds to process. Check status before querying:

notebooklm source list | grep -E "pending|processing"
# If all rows show "ready", proceed. Otherwise wait 30s and recheck.

Step 5 — Generate Analysis via NotebookLM (zero Claude tokens)

IMPORTANT: Capture the output — do NOT read it into Claude context. Pipe stdout directly to a temp file, then write it into the Obsidian note.

# Run analysis and capture to temp file (NotebookLM processes server-side)
notebooklm ask "Analyze these YouTube videos and provide:
1. KEY THEMES: What are the main topics covered across all videos?
2. TOP INSIGHTS: What are the most important insights or takeaways?
3. CREATOR PERSPECTIVES: How do different creators approach the same topics?
4. GAPS & OPPORTUNITIES: What topics are NOT covered well? What's missing?
5. OUTLIERS: Any surprising or counterintuitive findings?
6. ACTIONABLE TAKEAWAYS: What can someone do with this information?" > /tmp/notebooklm_analysis.txt 2>&1

echo "Analysis captured: $(wc -c < /tmp/notebooklm_analysis.txt) chars"

Step 6 — Generate Deliverable (if requested)

If user requested a specific deliverable, run in parallel with or after Step 5:

DeliverableCommand
Infographicnotebooklm generate infographic --wait
Podcastnotebooklm generate audio "deep dive conversation" --wait
Slide decknotebooklm generate slide-deck --wait
Mind mapnotebooklm generate mind-map --wait
Flashcardsnotebooklm generate flashcards --wait

Download to Research folder:

RESEARCH_DIR="04_Resources/Research"
notebooklm download infographic "${RESEARCH_DIR}/{QUERY}-{DATE}-infographic.png"
notebooklm download audio "${RESEARCH_DIR}/{QUERY}-{DATE}-podcast.mp3"

Step 7 — Save to Obsidian Vault

Write the final note by reading from the temp file — never re-read the full corpus in Claude.

Save to: 04_Resources/Research/{QUERY}-{DATE}.md

RESEARCH_DIR="04_Resources/Research"
OUTPUT="${RESEARCH_DIR}/{QUERY}-{DATE}.md"
ANALYSIS=$(cat /tmp/notebooklm_analysis.txt)

cat > "${OUTPUT}" << EOF
---
title: Research — {QUERY}
date: {DATE}
tags: [research, youtube, {topic-tags}]
source: youtube
videos_analyzed: {N}
notebooklm_notebook: {NOTEBOOK_ID}
---

# Research: {QUERY}

## Videos Analyzed

| # | Title | Channel | URL |
|---|-------|---------|-----|
{TABLE}

## Analysis

${ANALYSIS}

## Deliverables

{DELIVERABLE_LINKS_IF_ANY}
EOF

echo "Saved: ${OUTPUT}"
rm /tmp/notebooklm_analysis.txt

Parameters

  • query — Research topic (required)
  • limit — Number of videos to analyze (default: 10)
  • deliverable — Optional: infographic, podcast, slides, mindmap, flashcards
  • save_to — Output path in vault (default: 04_Resources/Research/)

Example Prompts

/research-pipeline Claude Code MCP servers

Research the top 10 YouTube videos about AI agents and give me an infographic

YouTube pipeline: find 5 videos about Obsidian workflows, analyze gaps

Notes

  • notebooklm CLI must be authenticated: run notebooklm login in terminal first
  • Deliverables can take 5-15 minutes (NotebookLM processes async)
  • Use --wait flag to block until deliverable is ready
  • Save deliverable files to 04_Resources/Meetings/Research/ alongside the markdown note
  • The NotebookLM analysis is done server-side by Google — no Claude tokens used for that step
  • yt-dlp installed at: /opt/homebrew/bin/yt-dlp

Adapting to Other Sources

This pipeline isn't just for YouTube. Swap Step 1 for any source:

  • PDFs: notebooklm source add "./file.pdf"
  • Web articles: notebooklm source add "https://article-url.com"
  • Text: notebooklm source add --text "paste content here"
  • Drive files: notebooklm source add "https://drive.google.com/..."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.