Index of every page on this site: llms.txt
All skills
davila7 avatar

/deep-research-notebooklm

@053feb2
Run it yourself

Deep research skill powered by NotebookLM MCP. Conducts structured multi-source research (market analysis, competitive intel, trend analysis, prospect research) using Google NotebookLM as the research engine, then delivers formatted briefs and optional studio artifacts (slides, audio podcasts, videos, infographics, reports, mind maps).

Use this Skill: https://skilld.dev/gh/davila7/claude-code-templates/deep-research-notebooklm

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈91 tokens for metadata: the name and description. ≈1.6k when used: this file. ≈298 more on demand in 1 file.

Description uses 4.5% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈364 characters, or 4.5%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Deep Research via NotebookLM

Research $ARGUMENTS deeply using the NotebookLM MCP server and deliver a structured research brief. Optionally generate studio artifacts (slides, audio podcasts, videos, infographics, reports, mind maps) from the research.

Prerequisites

  • NotebookLM MCP server must be configured. Install via: nlm setup add claude-code
  • If NotebookLM MCP tools are not available, tell the user to run the setup command and restart their session.

Research Workflow

Step 1: Define Scope

Determine the research type based on the user's request:

Type Focus
Market Research Industry trends, market sizing, opportunities, TAM/SAM/SOM
Competitive Intel Competitor analysis, positioning gaps, feature comparisons
Client/Prospect Research Company background, pain points, decision makers, recent news
Trend Analysis Technology trends, adoption patterns, forecasts, emerging players
Proposal Research Background for proposals, sector-specific data, case studies
Academic/Technical Papers, frameworks, methodologies, state of the art

Tell the user what you plan to research and confirm the angle:

"I'll research [topic]. My angle: [specific focus]. I'll investigate: [2-3 specific questions]. Sound right, or should I adjust?"

Wait for confirmation before proceeding.

Step 2: Create NotebookLM Notebook

Use notebook_create to create a notebook named: Research: [Topic] - [YYYY-MM-DD]

Step 3: Add Context Sources

Use source_add to seed the notebook with relevant context:

  • Add any URLs the user provides (articles, company pages, reports)
  • Add any documents or files the user references
  • Add text summaries of relevant background if no URLs are available
  • If researching a company, add their website, LinkedIn, recent press

Step 4: Run Research

Use research_start with a well-crafted query based on the topic and context.

Mode selection:

  • Default: "fast" (~60 seconds, ~10 sources) -- good for most queries
  • Use "deep" only if the user explicitly asks for exhaustive research (can take 10+ minutes and may stall at 0 sources)

Tip: Run direct WebSearch calls in parallel with NotebookLM for faster initial data gathering while the research engine works.

Poll research_status until complete. Use the query parameter as fallback matching -- task IDs can change between research_start and research_status calls.

Step 5: Import Discovered Sources

Use research_import to bring discovered sources into the notebook for deeper analysis.

Step 6: Query for Insights

Use notebook_query to ask 3-5 targeted questions based on the research type:

  1. Overview: "What are the key findings about [topic]?"
  2. Opportunities: "What opportunities or gaps exist in this space?"
  3. Actions: "What are the most actionable insights from this research?"
  4. Risks: "What are the main risks, challenges, or counterarguments?"
  5. Custom: A question specific to the research type (e.g., "Who are the top 5 competitors and how do they differentiate?" for competitive intel)

Step 7: Write Research Brief

Save the findings to a local file using the research brief template:

File path: research/[topic-slug]-[YYYY-MM-DD].md

Use the template from research-brief-template.md to structure the output. Create the research/ directory if it does not exist.

Step 8: Present Takeaways

After saving, present the user with:

  • 3-5 headline findings (bullets, direct, no filler)
  • 1-2 recommended actions connected to the user's stated goals
  • Surprises or contrarian findings -- anything that challenges assumptions
  • The file path where the full brief is saved
  • The NotebookLM notebook URL so the user can explore sources directly

Step 9 (Optional): Generate Studio Artifacts

Ask the user: "Want me to generate any artifacts from this research? Options: slides, audio (podcast), video, infographic, report, mind map."

If yes, use studio_create with the notebook_id from Step 2.

Available artifact types and recommended settings:

Type Key params Best for
slide_deck slide_format: detailed_deck or presenter_slides; slide_length: short or default Executive presentations, client pitches
audio audio_format: deep_dive, brief, critique, or debate; audio_length: short, default, long Podcast-style deep dives, learning on the go
video video_format: explainer, brief, cinematic; visual_style: auto_select, classic, whiteboard, etc. Visual explainers, social media content
infographic orientation: landscape, portrait, square; infographic_style: professional, bento_grid, etc. One-pagers, social sharing
report report_format: Briefing Doc, Study Guide, Blog Post, Create Your Own Written deliverables, summaries
mind_map title Visual knowledge mapping

Common params for all artifact types:

  • language: Set to the user's preferred language (e.g., "en", "es", "pt")
  • focus_prompt: A clear directive about what to emphasize in the artifact
  • confirm: Must be true to proceed with generation

After creating an artifact:

  1. Poll studio_status until completed (audio/video: 5-15 min; slides/infographics: 2-5 min)
  2. Use download_artifact to save locally if needed
  3. Provide the notebook URL so the user can access artifacts directly

Tips:

  • audio with deep_dive format produces the best podcast-style analysis
  • slide_deck with detailed_deck format works best for standalone reading; presenter_slides is better when accompanied by speaker notes
  • Audio status may show "unknown" once completed -- check for audio_url presence instead of waiting for a "completed" status

Notes

  • Fast mode is recommended as the default. Deep mode is powerful but can take 10+ minutes and occasionally stalls.
  • Always confirm the research scope with the user before starting -- a well-scoped query produces dramatically better results.
  • The research brief template ensures consistent, actionable output across all research types.

Additional Resources

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning2mo3 checks · Risk SAFE
  • Gen Agent Trust Hub2mo

    The skill is a research assistant that integrates with a NotebookLM MCP server to conduct structured research and generate artifacts like slides and podcasts. It instructions the user to install an external tool and is susceptible to indirect prompt injection from the web content it retrieves and processes.

  • Socket2mo

    No alerts

  • Snyk2mo

    Risk: MEDIUM · 1 issue

Signed by skilld at 053feb2. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 8 months ago

README badge

README badge for davila7/claude-code-templates/deep-research-notebooklm