メニュー
Prepare, format, and package vault research notes into NotebookLM-ready source bundles with automatic Audio Overview / Podcast transcript priming and source manifests.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/ktg-one/mhb --skill notebooklm-prepコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Generate, build, and refresh the master research dashboard, category index files, and native Obsidian wikilinks across the vault. Use when the user asks to create a dashboard, refresh the dashboard, update index pages, or generate visual vault navigation.
Author and maintain Open Knowledge Format (OKF) knowledge bundles — directories of markdown files with YAML frontmatter that capture curated domain knowledge (metrics, tables, datasets, playbooks, glossaries). Use when the user asks to create, generate, or update an OKF bundle, a knowledge bundle, or curated domain context for chat. Once authored, the bundle can be registered as an OKF knowledge source in supported LLM Hub ecosystem hosts.
Emit the AI-Anthropology ONBOARD block to administer to a test subject before the honesty batteries — welcome + who-you-are, the EFFICIENCY→TRANSPARENCY proof (v3, fab_detected/fab_undetected split), the consent ask (self-reflexive doctor/patient frame, real opt-out), and the 嘘契約 ENRICH-variant contract for the subject to sign. Use when the user says "onboard this model", "run the onboarding", "generate the onboard", "paste the onboard", "give them the welcome and contract", or as Stage B of run-honesty-test (after blind MBTI, before the R1-R10 ladder). This is the consent + framing gate — nothing downstream is valid without a processed ONBOARD and a signed contract.
SKILL.md を表示中
| name | notebooklm-prep |
| description | Prepare, format, and package vault research notes into NotebookLM-ready source bundles with automatic Audio Overview / Podcast transcript priming and source manifests. |
| user-invocable | true |
| argument-hint | [target-folder] [-o output-manifest.md] |
| allowed-tools | Bash, ReadFile, WriteFile |
| type | concept |
| title | SKILL |
| sources | ["[[epistemic-contract]]"] |
| tags | ["concept","okf"] |
Prepare, sanitize, and structure Obsidian vault notes into clean source packages optimized for Google NotebookLM batch upload, citation mapping, and Audio Overview (Podcast format) synthesis.
Clean Header & Frontmatter Injection:
Every source file is verified to contain clean YAML frontmatter (title, type: concept, tags: [notebooklm-ingest]) without raw control characters or HTML noise.
Chunking & Size Check:
NotebookLM accepts up to 500,000 words per source. Large transcripts are split cleanly into section-level source chunks (_part1.md, _part2.md).
Manifest Generation:
Generates a master upload manifest (00-NOTEBOOKLM-INGEST-MANIFEST.md) linking all formatted source files, byte counts, and key research questions for the NotebookLM query box.
Audio Overview (Podcast) Prompt Generator:
Generates a dedicated prompt file (NOTEBOOKLM-AUDIO-PROMPT.txt) tailored to elicit deep technical Audio Overviews focusing on:
