# Content migrations

Also follow [content working agreements](https://github.com/pufanyi/blog/blob/main/content/AGENTS.md), including image conversion, citation handling, and preserving author-owned prose.

- When migrating legacy posts, add spaces between Chinese text and Latin words, numbers, or inline math without changing the formulas or prose. Keep original code blocks verbatim; typography cleanup applies to prose only.

- For legacy Jekyll imports, replace Liquid includes with Markdown or native MDX and remove theme-only front matter. Keep original exports and ZIP archives outside `content/posts`; files not excluded by the asset rules are published.

- For Typst imports, translate math into standard TeX, preserve display equations and heading relationships, and resolve `@key` citations through a post-local `references.bib`. Keep the original `.typ` outside `content/posts`, since the asset copy rules would otherwise publish it. When inserting MDX with JavaScript `replace`, use a replacement callback: replacement strings turn `$$` display-math fences into single dollar signs.

- For Hexo imports, remove `<!--more-->` and replace plugin tags with equivalent embeds, such as a PDF iframe with a direct link. Apart from the prose spacing above, preserve original prose, math placement, links, and code; further editorial changes require a separate request. Resolve relative images in the asset folder named after the source post; prefer original PNGs over derived WebP files when converting to AVIF. Download links (PDF, ZIP, source code, and text) use `/posts/<slug>/<file>`; only image paths are rewritten automatically. Copy only referenced downloads and preserve their bytes. Update migrated cross-post links to `/blog/<slug>` and regenerate heading fragments with `slugifyHeading`; legacy Hexo anchors can differ in case and punctuation. Only level-two and level-three headings receive generated IDs; add an explicit ID when preserving links to deeper headings.