notecraft

name	notecraft
description	Google NotebookLM automation. Create notebooks, add sources, generate podcasts/videos/slides/flashcards, chat, deep research. Activates on intent like create a podcast, research topic, summarize URL.
user-invocable	true
allowed-tools	Bash, Read, Write
argument-hint	[research|podcast|analyze|chat] [args...]

NotebookLM

Automate Google NotebookLM via the notebooklm CLI (single static Go binary, assumed on $PATH). Generate audio podcasts, reports, slides, quizzes, videos, infographics, data tables, flashcards, analyze content, manage notebooks, and chat.

Setup

notebooklm --version              # Verify binary is on PATH
notebooklm export-session         # One-time: opens Chrome, log in to Google
notebooklm list --transport auto  # Verify

Session is saved to ~/.notebooklm/session.json by default. Override with --home <dir> or NOTEBOOKLM_HOME=<dir>.

If a browser isn't available on the host (CI, server), capture cookies elsewhere and bootstrap:

notebooklm import-cookies ./cookies.txt    # Netscape format, or Firefox/Safari profile path
notebooklm import-session ./session.json   # Or paste JSON inline
notebooklm refresh-session                 # Force a token refresh
notebooklm session-status                  # Cookie expirations + remaining validity

When This Skill Activates

Recognize requests like:

• "Create a podcast about [topic]"
• "Research [topic] in depth"
• "Summarize this URL"
• "Generate a report / study guide / blog post"
• "Generate flashcards / slides / quiz"
• "Create an infographic / data table / video"
• "Chat with my notebook"
• "What notebooks do I have?"
• "Add this PDF to my [company/topic] notebook"
• "Attach this URL to notebook X"
• "Create a new notebook called X" / "Rename notebook X to Y"
• "List / delete / rename / refresh / summarize sources in notebook X"
• "Add / list / update / delete notes in notebook X"
• "Make notebook X public" / "Share notebook X with [e-mail]"
• "Start web/deep research on topic Y" / "Import the research results into notebook X"
• "Describe / summarize notebook X" (AI-generated overview + suggested prompts)
• "Show / dump the raw text of source Y"
• "Delete this artifact" / "Revise slide N of this deck to say ..."
• "Export this report to Google Docs" / "Export this data table to Google Sheets"
• "Switch chat to learning-guide style / shorter responses" (configure chat)

CLI Commands

All commands use --transport auto for headless mode (Go uses native uTLS, no subprocess needed).

Task	Command
List notebooks	notebooklm list --transport auto
Notebook details	notebooklm detail <id> --transport auto
Create empty notebook	notebooklm create [title] --transport auto
Rename notebook	notebooklm rename <id> "New Title" --transport auto
Add source to existing notebook	notebooklm source add <id> --transport auto --file ./paper.pdf
Add URL to existing notebook	notebooklm source add <id> --transport auto --url "https://..."
List sources	notebooklm source list <id> --transport auto
Delete source(s)	notebooklm source delete <source-id...> --transport auto
Rename source	notebooklm source rename <notebook-id> <source-id> "New Title" --transport auto
Refresh source (re-fetch URL/Drive)	notebooklm source refresh <notebook-id> <source-id> --transport auto
Source summary (AI)	notebooklm source summary <source-id> --transport auto
Delete notebooks	notebooklm delete <id...> --transport auto
Chat	notebooklm chat <id> --transport auto --question "..."
List notes	notebooklm note list <id> --transport auto
Create note	notebooklm note create <id> --transport auto --title "..." --content "..."
Update note	notebooklm note update <id> <note-id> --transport auto --title "..." --content "..."
Delete note(s)	notebooklm note delete <id> <note-id...> --transport auto
Share status (collaborators + public state)	notebooklm share status <id> --transport auto
Public link on/off	notebooklm share public <id> --transport auto --enable (or --disable)
Invite collaborator	notebooklm share invite <id> --transport auto --email [e-mail] --permission viewer
Start research	notebooklm research start <id> --transport auto --query "..." --mode fast (or deep)
Poll research results	notebooklm research status <id> --transport auto
Import research into notebook	notebooklm research import <id> <research-id> --transport auto
Notebook AI summary + prompts	notebooklm describe <id> --transport auto [--json]
Source raw indexed text	notebooklm source content <source-id> --transport auto [--json]
Delete studio artifact(s)	notebooklm studio delete <artifact-id...> --transport auto
Revise slide deck	notebooklm studio revise <artifact-id> --transport auto --slide "0:Tighten intro" --slide "3:Add chart"
Export artifact to Docs/Sheets	notebooklm studio export <notebook-id> <artifact-id> --transport auto --format docs --title "Q2 Brief"
Configure chat goal/length	notebooklm chat configure <id> --transport auto --goal learning_guide --response-length longer
Chat with citations	notebooklm chat <id> --transport auto --question "..." --with-citations
Chat scoped to sources	notebooklm chat <id> --transport auto --question "..." --source-ids a,b,c
Podcast from URL	notebooklm audio --transport auto --url "https://..." -o /tmp/audio -l en
Podcast (debate, short)	notebooklm audio --transport auto --topic "AI" -o /tmp/audio --format debate --length short
Report (study guide)	notebooklm report --transport auto --url "https://..." -o /tmp/report --template study_guide
Report (custom)	notebooklm report --transport auto --url "https://..." -o /tmp/report --template custom --instructions "Write a SWOT analysis"
Slides	notebooklm slides --transport auto --url "https://..." -o /tmp/slides --format presenter
Video	notebooklm video --transport auto --url "https://..." -o /tmp/video --format explainer --style whiteboard
Quiz	notebooklm quiz --transport auto --url "https://..." -o /tmp/quiz --difficulty medium
Flashcards	notebooklm flashcards --transport auto --url "https://..." -o /tmp/flashcards
Infographic	notebooklm infographic --transport auto --url "https://..." -o /tmp/infographic --style professional
Data table	notebooklm data-table --transport auto --url "https://..." -o /tmp/table --instructions "Compare by category"
Analyze content	notebooklm analyze --transport auto --url "https://..." --question "Summarize"
Diagnose issues	notebooklm diagnose
Import session	notebooklm import-session <file-or-json>
Refresh tokens	notebooklm refresh-session

Source options (shared by all generation commands)

--url <url>              Source URL
--text <text>            Source text
--file <path>            Local file (pdf, txt, md, docx, csv, pptx, epub, mp3, wav, etc.)
--topic <topic>          Research topic (web search)
--research-mode <mode>   fast | deep (default: fast)

Generation command options

Command	Key options
audio	--format (deep_dive/brief/critique/debate), --length (short/default/long), --instructions, -l
report	--template (briefing_doc/study_guide/blog_post/custom), --instructions, --language
video	--format (explainer/brief/cinematic), --style (auto/classic/whiteboard/kawaii/anime/watercolor/retro_print), --instructions, --language
quiz	--instructions, --quantity (fewer/standard), --difficulty (easy/medium/hard)
flashcards	--instructions, --quantity, --difficulty
infographic	--orientation (landscape/portrait/square), --detail (concise/standard/detailed), --style (sketch_note/professional/bento_grid), --instructions, --language
slides	--format (detailed/presenter), --length (default/short), --instructions, --language
data-table	--instructions (describe table structure), --language

All generation commands require -o, --output <dir>.

Transport flag

--transport auto      Best available non-browser (recommended; uses native uTLS)
--transport http      Native uTLS HTTP/2 with Chrome fingerprint
--transport curl      curl-impersonate subprocess (requires binary on PATH)
--transport browser   Headed/headless Chrome via rod (requires --profile or --chrome-path)

Proxy flags (shared)

--proxy <url>         Generic proxy URL (any scheme)
--socks5-proxy <hp>   SOCKS5 host:port (scheme prepended)
--http-proxy  <hp>    HTTP host:port
--https-proxy <hp>    HTTPS host:port

Env-var fallback: SOCKS5_PROXY, HTTP_PROXY, HTTPS_PROXY, ALL_PROXY (lowercase variants also honored).

Multi-Account

notebooklm --home ~/.notebooklm-work list --transport auto
# or
NOTEBOOKLM_HOME=~/.notebooklm-work notebooklm list --transport auto

Autonomy Rules

Run automatically: list, detail, describe, diagnose, session-status, source list, source summary, source content, note list, share status, research status

Ask before running:

• Generation commands (long-running, creates notebook)
• create, rename (modifies notebook)
• source add / source rename / source refresh / source delete (modifies user's notebook)
• note create / note update / note delete (modifies notebook content)
• share public / share invite (changes visibility / invites collaborators — visible to others)
• research start / research import (kicks off long work / writes new sources)
• studio delete (irreversible artifact deletion)
• studio revise (creates a new slide-deck artifact; counts against daily quota)
• studio export (writes a new Google Doc/Sheet in the user's Drive)
• chat configure (changes notebook-scoped chat goal/length, persists server-side)
• delete (irreversible)
• refresh-session (rewrites session file)

Common Workflows

Quick podcast from URL

notebooklm audio --transport auto --url "https://example.com/article" -o ./output -l en

Research a topic (deep, debate format)

notebooklm audio --transport auto --topic "quantum computing" --research-mode deep -o ./output --format debate

Generate a study guide

notebooklm report --transport auto --url "https://example.com/paper.pdf" -o ./output --template study_guide --instructions "Focus on key formulas"

Generate slides from a topic

notebooklm slides --transport auto --topic "machine learning basics" -o ./output --format presenter --length short

Quiz from an article

notebooklm quiz --transport auto --url "https://example.com/article" -o ./output --difficulty hard --quantity standard

Analyze and ask questions

notebooklm analyze --transport auto --url "https://example.com/paper.pdf" --question "What are the key findings?"

Chat with existing notebook

notebooklm list --transport auto
notebooklm chat <notebook-id> --transport auto --question "Summarize the main points"

Chat with per-citation metadata (JSON output)

notebooklm chat <notebook-id> --transport auto --question "What does the paper claim?" --with-citations

Returns { text, threadId, responseId, citations: [{ index, sourceId, relevance, charStart, charEnd, excerpt, chunkId }] }. Use this when downstream code needs to render or verify per-sentence citations.

Add new material to an existing notebook

notebooklm source add <notebook-id> --transport auto --file ./new-report.pdf
notebooklm source add <notebook-id> --transport auto --url "https://static.cninfo.com.cn/finalpage/2026-04-16/1225107391.PDF"

Use this when the user wants to extend a long-running topic/company notebook with a fresh document rather than creating a new throwaway notebook.

Create a notebook, add a source, take a note

ID=$(notebooklm create "Q2 Earnings" --transport auto)
notebooklm source add "$ID" --transport auto --url "https://example.com/earnings.pdf"
notebooklm note create "$ID" --transport auto --title "Key takeaways" --content "Revenue +18% YoY..."

create prints the new ID to stdout; status messages go to stderr, so $(...) capture is clean.

Deep research → import as sources

notebooklm research start <notebook-id> --transport auto --query "enterprise AI ROI" --mode deep
# capture the printed researchID, then:
notebooklm research status <notebook-id> --transport auto    # polls, prints JSON
notebooklm research import <notebook-id> <research-id> --transport auto

status blocks until results are ready (default ~120 s timeout). import re-polls before writing so you can run it directly once start returns.

Share a notebook

notebooklm share status <id> --transport auto                       # current visibility
notebooklm share public <id> --transport auto --enable              # anyone with link
notebooklm share invite <id> --transport auto \
  --email [e-mail] --permission editor --notify     # direct invite

Triage a notebook before opening it

notebooklm describe <id> --transport auto                # AI overview to stdout + suggested prompts on stderr
notebooklm source list <id> --transport auto             # what's inside
notebooklm source content <source-id> --transport auto   # raw text dump for grep/diff

describe returns the notebook summary plus a few seed chat prompts. source content is the raw indexed text — useful for scripted search without going through chat.

Revise a slide deck after the first draft

notebooklm studio revise <artifact-id> --transport auto \
  --slide "0:Tighten the intro to two sentences" \
  --slide "3:Replace the bar chart with a timeline"

Creates a NEW slide-deck artifact with the instructions applied. The original is untouched. Use studio delete <old-artifact-id> afterwards if you want to discard the first draft.

Export a report / data table to Google Drive

notebooklm studio export <notebook-id> <artifact-id> --transport auto \
  --format docs   --title "Q2 Strategy Brief"        # report → Google Docs
notebooklm studio export <notebook-id> <artifact-id> --transport auto \
  --format sheets --title "Competitor Matrix"        # data table → Google Sheets

Prints the created document URL to stdout.

Tune the chat for a notebook

notebooklm chat configure <id> --transport auto \
  --goal learning_guide --response-length longer
# or a custom system prompt:
notebooklm chat configure <id> --transport auto \
  --goal custom --custom-prompt "Answer like a sceptical reviewer; cite sources."

Persists server-side, so subsequent chat calls (and the web UI) inherit the new style until you change it again.

Error Handling

Error	Action
"No session available"	notebooklm export-session (or import-cookies / import-session on headless hosts)
"Session expired"	Auto-refreshes via uTLS-fingerprinted client; if it still fails, re-export
"in-body UNAUTHENTICATED (code 16)"	Transport detected stale at/bl; refresh is automatic — only act if it loops
"Quota exceeded"	Daily limit hit — wait or upgrade
"Rate limited"	Wait a few minutes, retry
"curl-impersonate binary not found"	Either drop --transport curl (default auto picks native uTLS) or install curl-impersonate
"Audio download returned login page"	Re-run notebooklm export-session to refresh domain-scoped cookies for CDN downloads

Known Limits

• Audio generation takes 5–10 minutes; video can take longer
• Daily generation limits per artifact type — no API to query remaining quota
• Studio artifact types are server-driven (get_studio_config) — Google may add/remove types anytime
• Session auto-refreshes on at/bl expiry (~1–2 h); long-lived cookies last weeks
• Research deep-mode imports 40+ URL sources; readiness threshold is "70% indexed or 30 sources, whichever is lower" — a handful of unreachable URLs is tolerated by design
• Generation RPC sometimes returns a task ID that differs from the final artifact ID; the poller falls back to any ready AUDIO/VIDEO artifact with a download URL