Doc2X Markdown split proof, 2026-10-10 (Asia/Shanghai)

Input is self-authored UTF-8 text. Images are synthetic resources. No user document is included.
Actual output files were downloaded from https://doc2x.noedgeai.com/zh-cn/tools/md-split/ after selecting the input through the file chooser.
h2.zip: H2 headings, keep frontmatter on; 2 MD files.
h2-no-frontmatter.zip: same H2 input with frontmatter off; 2 MD files.
token-100.zip: reference tokens, target 100, keep frontmatter on; 6 MD files. Whole blocks may exceed this target.
outline-second.zip: outline extraction of the second chapter, keep frontmatter on; 1 MD file.
chapter-two.md: selected second chapter download, with one original header.

The output ZIP files contain Markdown only. We manually included input/images/chart.png in this verification package; the splitter did not copy or collect it. After extracting a chapter ZIP, copy the image directory into the relative location expected by that chapter.
Concatenating the H2 or reference-token files in sequence reproduces this input. The frontmatter-off output reproduces the body after the YAML header. Check PREFACE-KEEP-001, CODE-FENCE-KEEP-001 and DOCUMENT-END-20261010.
Rendering remains editor-dependent. Cross-chapter link/footnote definitions and image files are not copied into each fragment automatically.
