菜单
Library OS — the persistent digital library system. Use when adding a book, deepening an existing review, extracting insights from handwritten notes / highlights / photos, or when the user wants to build their own library on their site. Handles the full workflow from capture to publish.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | library-os |
| description | Library OS — the persistent digital library system. Use when adding a book, deepening an existing review, extracting insights from handwritten notes / highlights / photos, or when the user wants to build their own library on their site. Handles the full workflow from capture to publish. |
The open-source Library Intelligence System. Captures everything you read into a permanent, intelligent, citation-ready knowledge library on your own website.
BookReview type + its sub-types are the interoperable core. Port to any Next.js, Astro, SvelteKit, or even static HTML.CAPTURE ──────► EXTRACT ──────► ENRICH ──────► PUBLISH
│ │ │ │
▼ ▼ ▼ ▼
Photo, notes, quotes[] continueReading[] /library/{slug}
Kindle export, chapters[] videos[] (live, SEO+AEO)
voice memo tldr, faq + schema
Step 1 — CAPTURE The user provides the raw signal: a book title, a handwritten note photo, a highlights export, a voice memo. If a photo, extract text via vision. If a file, read it.
Step 2 — EXTRACT
Run /library-add to create the baseline entry (title, author, categories, 5 key insights, best-for, FAQ, TL;DR, cover). Then run /library-deepen to populate quotes and chapters via the book-distiller subagent.
Step 3 — ENRICH
Run /library-research to populate continueReading (external books with why-they-pair rationale) and videos (talks, interviews, lectures).
Step 4 — PUBLISH
Commit, sync to production repo, ship. Page goes live at /library/{slug} with the full deep-dive hub: TL;DR, TOC, insights, quotes, chapters, FAQ, continue-reading, videos, related-own-book, Amazon CTA.
| Command | Purpose | Writes to |
|---|---|---|
/library-add | Create a new entry from just a title | data/book-reviews.ts, public/images/library/ |
/library-deepen | Add quotes + chapters to existing entry | data/book-reviews.ts |
/library-research | Add continueReading + videos | data/book-reviews.ts |
Run in order. Each is idempotent — running /library-deepen twice just refines the existing quotes/chapters, it does not duplicate.
book-distiller — the extraction specialist. Given a book + (optional) source text, returns structured quotes[] and chapters[] matching the exact type shape. Delegated to by /library-deepen so the main conversation stays thin.
See .claude/agents/book-distiller.md for the agent prompt.
Located at app/books/types.ts:
interface BookReview {
// Identity
slug, title, author, coverImage, rating, reviewDate, categories, readingTime
// Core content (required)
keyInsights: string[] // exactly 5
bestFor: string[] // 3-4
// AEO core (optional, strongly recommended)
tldr?: string
faq?: Array<{q, a}>
publicationYear?: number
// Deep-dive hub (optional — the progressive depth layer)
quotes?: BookQuote[] // curated passages
chapters?: BookChapterSummary[] // number, title, keyIdea, summary
continueReading?: RelatedReadingItem[] // with reason, not algorithm
videos?: BookVideo[] // kind, duration, description
// Wiring
amazonUrl?: string
relatedBook?: string // slug from booksRegistry — links to our own books
hasCover?: boolean // gates the next/image render
}
Books without the deep-dive fields render as review-depth. Books with all fields render as full knowledge hubs. One template, progressive depth.
The intelligence is in the schema + workflow, not in any specific AI tool. The same pattern works with:
data/book-reviews.ts.claude/commands/library-*.md as plain instructions, execute the stepsSee docs/cross-ai-guide.md in the library-os repo for the prompt-paste versions.
The template is Next.js App Router (frankx.ai uses this). Adaptations:
app/library/ to src/pages/library/, use .astro for layout. Data schema unchanged.routes/library/+page.svelte, data unchanged.BookReview[] becomes frontmatter YAML in content/library/*.md.The spine — data/book-reviews.ts as a typed array + the JSON-LD schema — is portable.
The library index (/library) shows:
reviewDate DESC (newest first)The detail page (/library/{slug}) renders, in order:
/librarytldr)quotes)chapters)faq)continueReading)videos)relatedBook matches a slug in booksRegistry)All section IDs are anchor-linkable with scroll-mt-24 for smooth jumps from the TOC.
continueReading.reason must state a connection, not a summary.covers.openlibrary.org/b/isbn/{ISBN}-L.jpg) is fair-use for review purposes. Never scrape Amazon product images.Library changes live in the FrankX dev repo but deploy via the production repo (frankxai/frankx.ai-vercel-website). Standard pattern:
app/library/, app/books/types.ts, data/book-reviews.ts, public/images/library/)origin/main/library → the live index/library/approach → the showcase page explaining the systemlibrary-os repo → open-source template + docsbook-publishing (for authoring own books)Run the weekly FrankX operations loop — triage the inquiry inbox, advance the CRM pipeline, sweep open PRs across repos, and surface the content plan. Use for "/ops", "run the ops loop", "weekly ops", "triage inquiries", "what needs my attention", or any business-operations check-in.
Turn a NASA Astronomy Picture of the Day (APOD) into a 30–60s vertical short — script, shot list, captions, and a rights line. Use when the user mentions APOD, 'astronomy picture of the day', or wants a daily space short/reel/TikTok from a NASA image.
Turn an astrophysics/space-science arXiv paper into a structured brief tuned for space audiences — adds object/mission context, instruments, and observational caveats. Use when summarizing an astro-ph paper, a space-science result, or a mission's published findings.
Layer an Arcanea-style cosmic-mythic framing over a real space object or event, keeping the astronomy intact and separable from the myth. Use when giving a nebula/planet/launch a mythic narrative for the Arcanea universe without distorting the science.
Turn a NASA/ESA image into a structured 'cosmic atlas' web page — MDX with hero, plain-language explainer, fact table, sources, and rights. Use when building an atlas/encyclopedia entry, a gallery page, or a catalog of space objects from NASA imagery.
Check usage rights for NASA/ESA/observatory space media and emit a correct attribution line. Use when reusing a space image/video, before publishing a short or atlas page, or when the user asks 'can I use this image' / 'what's the credit' for NASA, ESA, Hubble, Webb, or observatory media.