notebooklm-cli-cookies_skill

Overview

This skill lets you search and answer questions over documents already uploaded to NotebookLM using the nlm CLI. It runs the exact nlm commands, verifies auth, and returns NotebookLM-based answers rather than answering from memory. Use it when you need authoritative responses sourced from a specific NotebookLM notebook.

How this skill works

The skill first checks that NOTEBOOKLM_MCP_CLI_PATH is set and that nlm login --check succeeds. It lists notebooks, resolves a notebook id or alias, then runs nlm notebook query with the user question and returns the stdout. If auth or resolution fails, it stops and asks for corrective steps instead of guessing.

When to use it

• You need a factual answer sourced from a NotebookLM notebook (not web or local files).
• You want to search or summarize documents already uploaded to NotebookLM.
• You have a notebook id or alias and want repeatable CLI queries.
• You invoked the skill via /nlm in Telegram and expect the raw args to be executed.
• You need the answer plus explicit notebook identity (title and id).

Best practices

• Ensure NOTEBOOKLM_MCP_CLI_PATH points to the auth storage directory before running queries.
• Prefer notebook aliases for repeated queries (create an alias like tai_lieu_dien).
• Resolve notebook titles to notebook_id from nlm notebook list --json; do not pass raw titles into query commands.
• If nlm login --check fails, refresh cookies outside of headless runtime and redeploy the secret. Do not attempt browser login in AWS.
• When a query returns empty or vague results, ask a refined follow-up question and include context to narrow the search.

Example use cases

• Ask "What is the warranty period for product X?" using the notebook alias tai_lieu_dien to get an answer sourced from that notebook.
• Summarize all sections related to safety procedures in a specific NotebookLM notebook by querying that notebook id.
• Resolve a customer's question by running nlm notebook query <alias> "Does the spec sheet mention feature Y?" and returning the CLI output.
• Use Telegram: send /nlm notebook query tai_lieu_dien "price of A9N61500" to execute the exact nlm args and get the stdout.