mdbook-llms
/mdbook-llms
Dodaje do książki mdBook pliki llms.txt, llms-full.txt i Markdown dla każdego rozdziału, dla modeli językowych.
- Autor
- ttdonovan ↗
- Repozytorium
- mdbook-llms ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 06.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 4
Brak sygnałów. Repozytorium zmienione 06.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Dodaje do książki mdBook pliki llms.txt, llms-full.txt i Markdown dla każdego rozdziału, dla modeli językowych.
Oryginalny opis (EN)
Add llms.txt, llms-full.txt and per-chapter Markdown output to an mdBook 0.5 book with the mdbook-llms renderer. Use when asked to make an mdBook readable by LLMs, add llms.txt to a book, configure [output.llms], or debug its output.
Jak działa
- Steps
- Gotchas
- Verify
Podgląd SKILL.md
---
name: mdbook-llms
description: Add llms.txt, llms-full.txt and per-chapter Markdown output to an mdBook 0.5 book with the mdbook-llms renderer. Use when asked to make an mdBook readable by LLMs, add llms.txt to a book, configure [output.llms], or debug its output.
---
# mdbook-llms
mdbook-llms is an mdBook 0.5 renderer. It writes `llms.txt` (an index of chapters), `llms-full.txt` (every chapter in one file) and a `.md` file per chapter. It ships with `mdbook-strip-frontmatter`, a preprocessor that keeps YAML frontmatter out of the HTML pages.
For every option and its default, read [USAGE.md](https://github.com/ttdonovan/mdbook-llms/blob/main/USAGE.md). This skill covers the steps and the gotchas.
The project's own docs are built with mdbook-llms and published for LLMs:
- [llms.txt](https://ttdonovan.github.io/mdbook-llms/llms.txt): index of every chapter, with descriptions. Fetch this first to find the right page.
- [llms-full.txt](https://ttdonovan.github.io/mdbook-llms/llms-full.txt): every chapter in one file, including usage, a worked example and the `Config` source.
## Steps
1. Check `mdbook --version` reports 0.5.x. Older versions are not supported.
2. Install both binaries with `cargo install --locked mdbook-llms`, and confirm `mdbook-llms` and `mdbook-strip-frontmatter` are on `PATH`.
3. Decide between the two layouts in `book.toml`:
- Alongside the HTML site (usual case): keep `[output.html]`, add `[output.llms]` and `[preprocessor.strip-frontmatter]`. Files go into `book/html/`.
- llms output only: list only `[output.llms]`. Files go into `book/`, and the strip-frontmatter preprocessor is not needed.
4. Set `site-url` under `[output.llms]` if links in `llms.txt` must be absolute. Without it, `output.html.site-url` is used, and without either, links are relative to the output root.
5. Optionally add a `description` to chapters through frontmatter. It is appended to that chapter's line in `llms.txt`:
```markdown
---
description: How to install the CLI
---
# Install
```
6. Run `mdbook build` and check the result (see Verify).
## Gotchas
- **Missing preprocessor**: if HTML is enabled without `[preprocessor.strip-frontmatter]`, frontmatter blocks are printed on the HTML pages. No `renderers` setting is needed; it already skips the llms renderer, which needs the frontmatter.
- **Frontmatter detection**: a block only counts if it opens the file with `---`, closes with `---`, and every line between looks like YAML. Anything else is treated as content.
- **Links to README.md chapters**: mdBook publishes `README.md` as `index.html` but does not rewrite links to it, so the HTML site links to a missing `README.html`. mdbook-llms fixes these links in its own output only. Avoid linking to `README.md` chapters if the HTML site matters.
- **mdbook serve**: it forces `output.html.site-url` to `/`, so when `[output.llms]` has no `site-url`, served llms links become root-relative (`/usage.md`). Use `mdbook build` for output you check or publish, and do not run serve and build into the same directory at once.
- **Hidden code lines**: lines mdBook hides in HTML (`# ` in Rust, `hidelines` prefixes elsewhere) appear in the Markdown output as ordinary code, marker removed. This is intended: agents get complete code.
- **Draft chapters**: chapters with no file (`[Title]()`) get no entry, and their children move up one level in `llms.txt`.
- **merge-into-html**: when HTML also runs, the renderer writes into `book/html/` and fails if no HTML output exists there. mdBook still creates an empty `book/llms/` directory, which can be ignored.
- **Optional files**: each `[[output.llms.optional-files]]` path is under `src/` and copied to the output, except `llms-full.txt`, which the renderer writes; a listed file that is missing fails the build. They are listed under `## Optional`, after every part.
- **Discovery links**: when merged, mdbook-llms edits the finished HTML pages to add `<link rel="alternate" type="text/markdown">` and `<link rel="describedby">` before `</head>`. Set `discovery-links = false` if another tool must own the HTML.
## Verify
- `llms.txt` exists in the output directory and lists every non-draft chapter, grouped under the `SUMMARY.md` part titles (or `## Contents` if there are none).
- Links in `llms.txt` use the expected prefix, and chapter pages exist at those paths (`README.md` chapters appear as `index.md`).
- No HTML page shows a frontmatter block.
- The build prints no `warning: mdbook-llms: broken link` lines. Each one names the generated file and the link; fix the source chapter, not the output.
- When merged, chapter HTML pages contain `rel="alternate"` and `rel="describedby"` links that resolve.
- `{{#include}}` directives are expanded in `llms-full.txt`.
- When unsure what correct output looks like, compare with the published [llms.txt](https://ttdonovan.github.io/mdbook-llms/llms.txt), which comes from a book with parts, nested chapters, a draft and descriptions.
Wersje
W bazie nie ma innych skilli o nazwie mdbook-llms.
Migawka bazy z 13.11.2026 · wersja paczki 092edd9c51ef (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100wiki-llms-txt
Generuje pliki llms.txt i llms-full.txt z dokumentacją projektu przyjazną dla modeli językowych, zgodnie ze specyfikacją llms.txt. - 100llms-txt
Tworzy lub poprawia plik llms.txt i llms-full.txt dla strony na podstawie jej prawdziwych podstron i dodaje go do kodu. - 100seo-llms-txt
Generuje, waliduje i audytuje pliki llms.txt dla widoczności w wyszukiwaniu AI: skanuje strukturę witryny i tworzy zgodny ze specyfikacją indeks w Markdown dla modeli językowych. - 100geo-fix-llmstxt
Generuje pliki llms.txt i llms-full.txt dla strony internetowej, aby poprawić jej czytelność i wykrywalność przez systemy AI. - 100create-llms
Tworzy plik llms.txt od zera na podstawie struktury repozytorium, aby modele językowe mogły odkrywać zawartość projektu. - 100llms-txt-writer
Pisze dokumenty dla systemów AI (llms.txt, llms-full.txt, strony FAQ, glosariusze) według standardu llms.txt, gdy jedynym czytelnikiem dokumentu jest system AI.