Reading with LLMs — llms.txt Quiz

1. What does this guidebook's llms.txt file contain, and what is it for?
   • A) The full Korean content in markdown, for pasting whole into a context window
   • B) An index of each content page's title, summary, and raw Markdown URL, for letting an LLM pick and fetch only the pages it needs
   • C) A collection of quiz answer keys, for automated grading
   • D) A bundle of diagram PNGs, for building slides

Show Answer

Answer: B) An index of each content page's title, summary, and raw Markdown URL, for letting an LLM pick and fetch only the pages it needs

Explanation: In the endpoints table, llms.txt is an index of each content page's group, title, summary, and raw Markdown URL, used for letting an LLM pick and fetch only the pages it needs. Each content link uses /llms/<locale>/<source path>.md, so the model reads the document source without a rendered page's sidebar or scripts. The full markdown content lives in the separate llms-full-ko.txt / llms-full-en.txt files. Quizzes and labs are not listed page by page; ## Optional carries one quiz-index link and one lab-index link per language.

2. In which languages are the full-content files published?
   • A) Korean only (llms-full-ko.txt)
   • B) Korean and English — two files (llms-full-ko.txt, llms-full-en.txt)
   • C) Five languages: Korean, English, Chinese, Japanese, and Spanish
   • D) A single llms-full.txt with every language merged together

Show Answer

Answer: B) Korean and English — two files (llms-full-ko.txt, llms-full-en.txt)

Explanation: The language-wide full files are llms-full-ko.txt and llms-full-en.txt. The index, manifest, section bundles and per-document Markdown are additional endpoints served under https://www.atomai.click/kubernetes-docs/. The Korean file is recommended for "Whole-book context or RAG indexing" and the English one for "English-language tools and pipelines".

3. How are llms.txt and full files produced for a successful site deployment?
   • A) The author re-uploads the three files by hand after every edit
   • B) A weekly news-digest workflow refreshes them
   • C) The site build generates them from the same source and deploys them together
   • D) The server converts the documents on the fly for each LLM request

Show Answer

Answer: C) The site build generates them from the same source and deploys them together

Explanation: The page states: "The site build generates them from the same source and deploys them together, so they always match the published content." They are build outputs of the deploy pipeline — not manual uploads and not on-demand conversions.

4. How are quiz documents handled in llms.txt and the full files?
   • A) Every quiz page appears in both the index and the full files, answer keys included
   • B) Only the quiz index page is linked (one link per language), under ## Optional in llms.txt; individual quiz pages and their answer keys are left out of both the index and the full files
   • C) Every individual quiz page is listed in the index, but the full files drop them
   • D) Quiz pages appear in the full files but are excluded from the index

Show Answer

Answer: B) Only the quiz index page is linked (one link per language), under ## Optional in llms.txt; individual quiz pages and their answer keys are left out of both the index and the full files

Explanation: The page says: "Quizzes appear in llms.txt only as a link to the quiz index page (one per language) under ## Optional; individual quiz pages (with their answer keys) are left out of both the index and the full files — answer keys don't belong in an LLM's context." So an LLM can discover that quizzes exist and where each language's quiz index lives, but neither the flat index nor the whole-book files ever carry a single quiz page or its answers. Lab guides differ: they are also only an index link in llms.txt, but their full text is included in the full files.

5. Given the size of the full files, which way of using them with an LLM does the page recommend?
   • A) Paste the multi-MiB full file into every prompt — it is the most accurate approach
   • B) Split the full file into several pieces and attach all of them every time
   • C) Point the model at the index and let it fetch specific pages — this works better with most tools
   • D) Hand over only the diagram viewer URLs and let the LLM infer the rest

Show Answer

Answer: C) Point the model at the index and let it fetch specific pages — this works better with most tools

Explanation: The "Size" note says the full files run several MiB and that "for most tools, pointing the model at the index and letting it fetch specific pages works better than pasting the whole book into one prompt." The first example prompt ("Read …/llms.txt, find the document … and summarize …") is exactly this pattern. The full files suit whole-file download-and-chunk uses such as RAG indexing.

6. Why is per-document chunking of llms-full-*.txt described as trivial?
   • A) Each document is stored as a separate ZIP entry
   • B) Every document is preceded by a separator block containing Source: <URL>
   • C) Each document is cut to a fixed 4,096 characters
   • D) The documents are stored one per element in a JSON array

Show Answer

Answer: B) Every document is preceded by a separator block containing Source: <URL>

Explanation: Under "Format notes", "every document is preceded by a separator block" — a dashed line, a Source: https://www.atomai.click/kubernetes-docs/en/core/01-cluster-architecture-style URL, and another dashed line. The RAG example's comment repeats it: "documents are separated by Source: <URL> blocks, so per-document chunking is trivial." The file is plain markdown text, not ZIP or JSON.

7. How does the page describe the way an LLM and a person each use the diagram information carried in the full files?
   • A) The LLM analyzes the PNG image directly while the person reads the alt text
   • B) The LLM reads the diagram's description (alt text) to understand it, while a person opens the interactive viewer URL and uses the Export menu to download PNG/JPEG/WebP, SVG, WebM, or a Share Card
   • C) Neither the LLM nor a person can get any diagram information from the full files
   • D) The LLM renders a WebM animation from the viewer URL and hands it to the person

Show Answer

Answer: B) The LLM reads the diagram's description (alt text) to understand it, while a person opens the interactive viewer URL and uses the Export menu to download PNG/JPEG/WebP, SVG, WebM, or a Share Card

Explanation: The "Diagrams are for people — the export links" section explains that because the full files carry each document's markdown verbatim, every diagram's alt text and its viewer URL (https://www.atomai.click/kubernetes-docs/archmaps/<name>.html) are in the text. "A text LLM uses the alt text and surrounding prose but cannot infer image-only nodes/edges reliably; a person opens the URL and uses the viewer's Export menu to download PNG/JPEG/WebP, a dual-theme SVG, a 6-second trace-animation WebM, or a 1200×630 Share Card." The per-item guidance lives in the Guidebook Roadmap's "Share a diagram" section.

8. What limit does the page place on diagram files exported from the viewer (PNG, SVG, WebM, Share Card)?
   • A) Exports may be used as formal evidence that the architecture was validated
   • B) Exports are communication assets, not evidence that an architecture was validated
   • C) Exports are for LLMs only and cannot be opened by people
   • D) Exports are the official release form of the document and replace the source HTML

Show Answer

Answer: B) Exports are communication assets, not evidence that an architecture was validated

Explanation: The section closes with: "Exports are communication assets, not evidence that an architecture was validated." Exporting a diagram serves sharing purposes such as LinkedIn posts or talk slides; the existence of the image says nothing about whether the architecture it depicts has been validated.