notebooklm-py-skill

该 Skill 指导使用本地 notebooklm CLI 操作 Google NotebookLM：先校验认证，再列出笔记本、添加网址/文件/文本来源、提问、等待来源就绪、查看产物与导出；要求优先使用显式笔记本 ID 与 --json 输出，并在删除、生成、下载等破坏性或耗时操作前先行确认。

suidge

热度 16.16

最近更新 2026-05-24 16:13:19

分类：知识依赖：代码执行 / 读写文件 / 第三方 API应用场景：笔记整理 / 研究 / 摘要总结 / 工作流

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add notebooklm-py-skill

Metadata

name: notebooklm-py-skill
description: Manage Google NotebookLM notebooks, sources, chats, artifacts, and exports through the notebooklm-py CLI with safe workflows.
homepage: https://github.com/Suidge/notebooklm-py-skill
license: MIT
metadata: {"openclaw":{"skillKey":"notebooklm-py-skill","emoji":"📓"}}

notebooklm-py-skill

Use this skill when the user asks to operate Google NotebookLM through the local notebooklm CLI: notebooks, sources, chat, research, notes, sharing, artifacts, or exports.

Core Rules

Verify auth with notebooklm auth check --test --json before any workflow that depends on account access.
Prefer explicit notebook IDs (-n NOTEBOOK_ID or --notebook NOTEBOOK_ID) in automation; avoid shared notebooklm use context in parallel workflows.
Prefer --json for commands whose output will be parsed, cited, or used as evidence.
Ask before destructive or long-running work: notebook delete, source delete/clean, artifact delete, generation, downloads, saved notes, and waits in the main conversation.
Put one-off prompt/source/output files under a temporary directory and clean them up unless the user asked to keep them.
Never run notebooklm doctor --fix without explicit confirmation.

Common Commands

notebooklm auth check --test --json
notebooklm list --json
notebooklm metadata -n NOTEBOOK_ID --json
notebooklm source list -n NOTEBOOK_ID --json
notebooklm source add "https://example.com" -n NOTEBOOK_ID --json
notebooklm source add ./file.pdf -n NOTEBOOK_ID --json
notebooklm source add - --type text --title "notes.md" -n NOTEBOOK_ID --json < /tmp/openclaw-attachments/notes.md
notebooklm ask "Question" -n NOTEBOOK_ID --json
notebooklm ask --prompt-file /tmp/openclaw-attachments/question.txt -n NOTEBOOK_ID --json
notebooklm artifact list -n NOTEBOOK_ID --json

Workflows

Query A Notebook

Run auth verification.
Locate the notebook with list --json, or use the provided ID/URL.
Ask with ask ... -n NOTEBOOK_ID --json.
If the answer is incomplete, ask focused follow-ups and synthesize.

Add Sources Then Query

Create or identify the notebook.
Add URL/file/text sources with source add ... --json.
If the next step depends on indexing, wait for returned source IDs with source wait SOURCE_ID -n NOTEBOOK_ID --json.
Ask or generate only after required sources are ready.

References

Load references/notebooklm-py-cli.md for command groups, 0.5.0 compatibility notes, artifact examples, and troubleshooting.