| name | notebooklm-artifact-bundle |
| description | Use when Codex needs to generate the repository's standard NotebookLM artifact bundle for a notebook, inspect artifact/task status, reuse the established audio/slide/video prompt texts, or upload generated audio back into NotebookLM to recover a transcript. |
NotebookLM Artifact Bundle
Use the repo-local NotebookLM workflow in .venv/bin/notebooklm. This skill is for repeated artifact generation and transcript recovery, not for generic NotebookLM brainstorming.
Working Rules
- Use the repo-local binary:
.venv/bin/notebooklm.
- Verify auth first with
.venv/bin/notebooklm auth check --test.
- Resolve the target notebook before generating artifacts. Accept either notebook ID or exact title.
- Submit the standard bundle first, then poll in batches instead of serially waiting on each task.
- Reuse the established prompt texts for audio, slide deck, and video unless the user explicitly asks for a different tone.
- When the user wants a transcript of generated audio, upload the MP3 back as a source, wait until it is ready, then export the indexed full text.
- Treat downloaded NotebookLM media as raw artifacts. Do not delete semantically named article images or remove markdown image embeds while generating or refreshing the bundle.
- Before committing an artifact refresh, inspect
git diff --name-status for deleted assets and git diff -- <draft> | rg '^-!\[' for removed image embeds.
- For
source fulltext, save with -o <path> instead of shell redirection. The terminal view can be a preview and may not contain the full transcript.
Standard Bundle
- Audio overview:
deep-dive, long
- Slide deck:
detailed
- Flashcards:
more, hard
- Quiz:
more, hard
- Video overview:
explainer
Reference Map
../../../workflows/notebooklm-artifacts/README.md: tracked canonical artifact-generation and transcript-recovery workflow.references/workflows.md: notebook resolution, standard bundle generation, polling, and transcript round-trip.references/prompts.md: canonical prompt files and the voice constraints they encode.
