Convert archived or specified articles into learning-oriented or interactive outputs: quizzes, flashcards, mind maps, podcast scripts, and study notes, prioritizing NotebookLM and falling back to local Markdown.
Customer-readable description
What this skill can do
Convert Markdown / vault articles or URLs into learning outputs:
- Quiz: multiple-choice questions + short-answer questions + answer explanations
- Flashcards: Anki-compatible double-sided cards
- Mind map: Mermaid / Markdown hierarchical structure
- Podcast script: conversational audio script
- Study notes: NotebookLM comprehensive summary
| Customer wants | Skill will do | Customer can see |
|---|---|---|
| Generate quiz | Local Markdown or NotebookLM quiz artifact | Number of questions, question type distribution, file path |
| Make flashcards | Anki-compatible Markdown or NotebookLM flashcards | Number of cards, file path |
| Mind map / podcast | Mermaid mind map / dialogue script Markdown | File path, number of nodes / word count |
| Upload to NotebookLM | Upload source files, record notebook id | notebook id / artifact path |
| Execution complete | Acceptance that the output structure is complete and the number of questions and answers is consistent | Completion receipt |
How customers use it
- Specify the source (URL / vault path / local Markdown) and the target type (quiz/flashcards/mind map/podcast, optional).
- Optional: specify provider (
local/notebooklm), number of questions, number of cards, difficulty. - For URL sources, clip and archive first, then convert.
Dependencies and installation
npx skills add soia-team/soia-open-skills -g -a '*' -s soia-pkm-transform-article-notebooklm -y
- Local solution: no additional dependencies (generates Markdown)
- NotebookLM (optional): see references/provider-notebooklm.md
Logs and completion receipt
完成：<一句话>。
日志摘要：
- source: <路径或 URL>
- artifact_type: quiz | flashcards | mindmap | podcast | notes
- provider: local | notebooklm
- content_mode: learning
- output: <文件路径或 notebook id>
- count: <题数 / 卡片数 / 节点数>
验证：
- 文件存在且可读
- 题目数 == 答案数（quiz/flashcards）
- 脑图 Mermaid 可解析（mindmap）
- prompt 已落盘可重跑
问题与下一步：
- <无 / provider 降级原因 / NotebookLM 失败目标列表>
Boundaries
- Default
learningmode: structural reorganization, extraction of core knowledge points, without compressing core key points. - quiz/flashcards: number of questions == number of answers == number of explanations; acceptance must ensure all three are consistent.
- prompt is written to disk at
outputs/transform/<YYYY>/<stem>/prompts/。 - NotebookLM is an optional provider; when missing, use local Markdown generation and do not stop the task.
- Record each failed NotebookLM artifact individually (must not be silently skipped).
Workflow
- Confirm the source → clip URLs first.
- Determine
artifact_type：quiz(default) /flashcards/mindmap/podcast/notes。 - Select provider: user-specified > configured > local Markdown.
- NotebookLM path: run health check first (references/provider-notebooklm.md), then upload source files, record notebook id, and generate the corresponding artifact.
- Local path: read references/prompt-notebooklm.md corresponding sections to generate the prompt, and output Markdown.
- Generate output to
outputs/transform/<YYYY>/<stem>/, prompt saved to disk. - Acceptance: complete structure, consistent counts, Mermaid parseable (if applicable).
- Receipt, NotebookLM failed targets listed item by item.
See references/prompt-notebooklm.md、references/notebooklm-test-matrix.md、references/quality-gates.md。