Summary
Add a real figure block for standalone images, captions, labels, and future figure-level metadata.
Right now the format spec only has inline image syntax:
FORMAT.md:48 — 
That is fine for lightweight inline images, but it is the wrong semantic shape for figures that need captions, numbering, layout controls, and cross-references.
Motivation
Standalone images usually behave like document blocks, not inline phrasing content.
A block form would make it possible to support:
- captions
#fig:... labels
- figure numbering
- cross-references like
[@fig:...]
- future figure metadata such as width/alignment/source attribution
- multiple images in one figure later if needed
Proposal
Keep plain  for simple inline usage, but add a block-level figure form for richer semantics.
A natural fit for Coflat is the existing fenced-div plugin architecture. For example:
::: {.figure #fig:architecture}

System architecture overview.
:::
Expected behavior
- Inline
 remains supported.
- A standalone
figure block renders as a block container.
- Caption text is editable as normal document content.
- Labels participate in numbering and cross-reference resolution.
- Rich/read/source behavior should stay coherent like other block plugins.
Why separate this from plain images
Treating every image as a block would make inline image usage awkward.
The better split is:
- inline image syntax for lightweight inline media
figure block for semantic document figures
Relevant context
Current format/spec only documents inline images and block tables/code/etc. There is no figure-level construct yet.
Summary
Add a real
figureblock for standalone images, captions, labels, and future figure-level metadata.Right now the format spec only has inline image syntax:
FORMAT.md:48—That is fine for lightweight inline images, but it is the wrong semantic shape for figures that need captions, numbering, layout controls, and cross-references.
Motivation
Standalone images usually behave like document blocks, not inline phrasing content.
A block form would make it possible to support:
#fig:...labels[@fig:...]Proposal
Keep plain
for simple inline usage, but add a block-levelfigureform for richer semantics.A natural fit for Coflat is the existing fenced-div plugin architecture. For example:
::: {.figure #fig:architecture}  System architecture overview. :::Expected behavior
remains supported.figureblock renders as a block container.Why separate this from plain images
Treating every image as a block would make inline image usage awkward.
The better split is:
figureblock for semantic document figuresRelevant context
Current format/spec only documents inline images and block tables/code/etc. There is no figure-level construct yet.