An Enscribe document opens with a metadata block. Here is the one at the top of this page:

<meta type=article>
  <title | Getting Started with Enscribe>
  <author | The Enscribe Project>
</meta>

<meta> carries the document’s identity — title, authors, and optionally a <date>, an <abstract>, affiliations, and the rest of the front matter a scholarly article needs. It is pinned to the very top because it describes the document rather than belonging to its body; Enscribe keeps such apparatus at the document’s edges, where structural processing can find it. The type=article attribute picks the article shape; type=book would instead expect chapters. After the metadata you simply write. A blank line starts a new paragraph, exactly as in Markdown. That part you know — so the rest of this guide is the parts you don’t.

A heading can be written three ways, and all three compile to the same thing — an eHTML <section> wrapping a <section-title>:

<# Methods #>           — sigil form: the fewest keystrokes
<section | Methods>     — canonical form: explicit; what the others become
# Methods               — Markdown form: familiar, and it works

If you are in the playground, try it: change one of this guide’s <# ... #> headings to # ... and back. The rendered heading does not move — because all three forms reduce to the same node before anything else in the pipeline sees them.

Why three? Because Enscribe separates the form you type from the form it becomes. Underneath every document is one canonical representation — semantic HTML — and the shorthand is just a faster way to reach it. The sigil <#> is the day-to-day default: it is the lightest to type and still carries everything a section needs. The named <section> is the explicit form, and it is exactly what you would read if you converted the finished HTML back to shorthand. The Markdown # is there so muscle memory works on day one. This guide writes its own headings in the sigil form — the line just above this paragraph was <# Sections come in three forms #>. Give a section an id when something will point back to it, by adding #name before a pipe: <# #sec:forms | Sections come in three forms #>. The three forms are not three parsers racing each other.¹ Their trade-offs are laid out in table 1 below.

Inside a paragraph, **bold** is bold, *italic* is italic, and backticks give inline code — all plain Markdown, unchanged. Enscribe does not reinvent the things Markdown already does well; it delegates them and adds only what is missing. The first addition you reach for in scholarly prose is mathematics. Inline math uses LaTeX dollar signs: typing $\varphi = (1 + \sqrt{5})/2$ sets the golden ratio, $\phi =(1+\sqrt{5})/2$, right in the run of text. Everything between the dollars is handed to KaTeX, the same engine many journals use.² Delegating the hard parts — math, syntax highlighting, citation formatting — is a deliberate stance: Enscribe owns the vocabulary, not the typesetting of every embedded language.

A result that deserves its own line gets the display form, written with the doubled-dollar tag <$$ | ... $$>. Here is the Gaussian integral — the kind of identity a methods section states once and then leans on:

Display math is centered and set larger than inline math, and like a figure or a table it can take an id (<$$ #eqn:gauss | ... $$>) so the prose can refer back to it. In the playground, try changing the exponent — the equation re-typesets on every keystroke.

Lists carry their own marker. Each item starts with <li> (or the sigil <->, or a markdown -), and an item runs until the next marker — so nesting is just a <list> inside an item, with no indentation to count:

• Gather the data

• Fit the model

  1. Choose the priors
  2. Run the sampler

• Write it up

The inner list there is <list ordered>, so it numbers while the outer stays bulleted. marker= (any CSS list-style-type), start, and reversed are there when you need them.

Tables are laborious in raw HTML and finicky in Markdown, and academic tables are mostly data. Enscribe lets you write the rows as CSV and supplies the table scaffolding — the element, the header row, the cells — for you. The source below produced table 1:

<table #tab:layers csv caption="..." |
Form,Keystrokes,Carries an id,Round-trips from eHTML
Canonical,most,yes,yes
...
>

The first line is the header; each line after it is a row. The caption attribute is rendered below the table, and the table number is added for you. Rendered, it is:

That table is the three-layer model made concrete: the same structure, three notations, one meaning. Try adding a row in the playground; the table grows to fit it.

A <figure> interrupts the text flow, carries an optional caption, and is numbered automatically — you never type the number. Figures, tables, code blocks, and the rest share this one frameable capability, so they all behave the same way: caption below, number folded in, an optional border. The image source can be a local path or a URL:

Because the figure carries an id, prose elsewhere can point at it by name rather than by a number you would otherwise have to keep correct by hand: see figure 1.

You have already met footnotes — the two asides above are <note | ... > tags, written immediately after the word they attach to, with no space. On the docs site they render as small markers you can hover to read in place, without losing your spot; in print they fall to the foot of the page or the end of the section. Same source, different surfaces.

Citations work against a bibliography you supply. The <data> block at the end of this page holds a <library> of BibTeX entries — apparatus like <data> and <config> goes at the document’s edges, by convention at the end. Nothing is fetched from disk, so the whole document stays self-contained and editable in the browser. You cite an entry by its key, and the formatting is not Enscribe’s job: it is delegated to citation-js and the CSL styles the rest of scholarly publishing uses, so switching from one journal’s style to another’s is a configuration change, not a rewrite.³ Markdown (Gruber, 2004) is the ancestor every lightweight markup language descends from, and Enscribe keeps its ease for prose while refusing its ceiling. HTML (WHATWG, 2024) is the rendering substrate underneath — already a complete, universal document format. And JATS (National Information Standards Organization, 2021), the archival XML the scholarly-publishing world runs on, is Enscribe’s downstream export target. Each marker is numbered as it appears, and the full, formatted list is gathered at the foot of the page.

You have seen two cross-references already: table 1 and figure 1. A <ref> names its target by a typed id — tab:layers, fig:pipeline, eqn:gauss — where the prefix says what kind of thing is being referenced. Enscribe resolves these in a pass after the whole document is parsed, so a reference can point forward or backward freely, and the rendered number (“Table 1”, “Figure 1”) is always the one the target actually received. Move the table, renumber nothing — the references follow.

The theorem family — <theorem>, <lemma>, <proof>, and friends — gives formal statements their own numbered, styled blocks, each on its own counter. A <definition> is the plainest member, and this guide has been leaning on one term that earns it:

Scroll back up. In a few dozen lines of source you wrote a document with sections in three notations, inline and display math, a captioned table and a captioned figure, two cross-references that found their own targets, hover-preview footnotes, three real citations with an automatically formatted bibliography, and a numbered definition — and the rendered result is self-contained, semantic HTML that a browser shows with no toolchain in the way. That is the whole pitch in one page: the ease of Markdown, the reach of HTML, and the apparatus of scholarly writing, in a single source file you can read, edit, and archive.

For the architecture behind the two-layer model — why eHTML is canonical, how the shorthand lifts to it, and where JATS fits — read the Design article. The Authoring Guide, covering every element in depth, and the eHTML Reference, cataloguing the output vocabulary, are the next documents to land on this site. Until then, the best next step is the one right beside this text: keep editing, and watch what changes.