Ok, figured it out. Protocol-Buffer in linguist is for highlighting Protocol-Buffer files that conform to the language specification.
The files that Defold outputs are text encoded protocol buffer messages and can be translated using protoc --encode and protoc --decode accordingly.
I don’t know that there will ever be a syntax for the text format available, as it is not a language itself, but a data packet not too dissimilar from the binary format that Protocol Buffers would normally pass messages around with.
If I mark them as JSON5 I get syntax highlighting because of the unquoted key names, however, the comment in the format is
# and does not highlight properly, as well as JSON5 expects a comma at the end of the line instead of a new-line, however this does not break the highlighter.
The trick here is that anything defined as “data” instead of a language, would not count against your stats, as data is not accumulated in github for a language score. So using JSON5 for now would get us halfway there, and unless someone creates a hiughlighter for text protobuf messages, nothing else really exists, and it would look like plain text.
Simply marking the types as ‘linguist-detectable=false’ is enough to remove them from the language stats, without attempting to pretty the output when viewed on github.