from notebooklm-skill157
End-to-end research-to-content pipeline: ingest sources into Google NotebookLM, run deep cited research, generate downloadable artifacts, and produce polished c
NotebookLM Research Agent automates source ingestion into Google NotebookLM, runs web/Drive research, extracts citation-backed answers, and generates downloadable artifacts (audio, video, slides, reports, quizzes, flashcards, mind maps, data tables). It then uses Claude to turn research outputs into polished content like blog posts, social threads, and newsletters. The repo also exposes an MCP server so the skills can be called as tools from Claude Code or other MCP-compatible clients.
Use this skill when you need source-based, cited research synthesized into publishable outputs. Typical triggers: "research this topic", "create a notebook about X", "turn these articles into a post", "generate a podcast from these sources", or "deep research on X". It's ideal for producing evidence-backed articles, show notes, slide decks, or datasets derived from collected sources.
Best with Claude Code and other MCP-capable tools (Claude, Claude Code, Cursor). The repo is designed to be used where an assistant can call CLI or MCP tools to manage notebooks and artifact pipelines.
NotebookLM Research Agent is a comprehensive research-to-content pipeline skill that ingests sources into Google NotebookLM, runs deep web research, generates 10 artifact types (audio podcast, video, slides, report, quiz, flashcards, mind map, data table, study guide, infographic), and produces polished content drafts via Claude. Scripts are well-structured CLI wrappers around notebooklm-py v0.3.4 with proper error handling, JSON output contracts, and clear arg parsing. All scripts exited cleanly showing usage help when invoked without required arguments in DRY_RUN mode — expected behavior, not a bug.
notebooklm-pypython-dotenvfeedparserffmpegpoppler (pdftoppm)Well-crafted skill with genuine utility. The 4-phase pipeline (Ingest→Synthesize→Create→Publish) is clearly documented and the scripts are production-quality. MCP server integration adds value for Claude Code/Cursor/Gemini users. Multilingual support (ZH-TW triggers) is a nice touch. Only minor concern: make_video.sh eval usage is mild risk, but acceptable since inputs are user-provided file paths, not remote data.
