I did something similar but used wiki syntax with a handful of extensions. The source content was stored in a specific directory layout/scheme and the site code generated the page hierarchy from that, storing it in an adjacent cache dir. Coming from a sysadmin background, it seemed very natural and easy to manage. Definitely the way to go.
One day I'll clean it up and offload it to github for others who are like-minded.
This sounds almost exactly like something I whipped up in the early 2000s [1]. Each post was its own directory (so I could just drop images in), and it used file system metadata for dates. It was based on Bash, ImageMagick, sed, m4, head, tr, and... XSLT (I swear it seemed like a good idea at the time!).
Believe it or not, the tool actually even worked with Cygwin after I switched to Windows (on my very first laptop).
One day I'll clean it up and offload it to github for others who are like-minded.