After dumping a script and formatting it, you should "sanity check" the resulting text file to make sure that it doesn't have obvious problems.

The purpose of sanity checking is to validate that the script isn't obviously broken, without checking the whole script against a playthrough of the game or something.

0:

Make sure the script doesn't use the 《》 characters on its own before you format ruby text with 《》. If it does, remove them or replace them with «».

1:

Look for formatting characters. Note that some of these characters have special meaning in regexes, so disable regexe search or escape the relevant characters.

The most common formatting characters are: @#$%^&*-=_+[]{}\|;:<>/

More characters that can be formatting are sometimes found in the script's content text instead of as formatting: ~`!()"',.

You should also look for alphanumeric ascii and make sure that any strings of it are in content text instead of filenames or formatting:

[a-zA-Z0-9]*

That's a regex that searches for any series of character consisting entirely of characters from the ranges a-z, A-Z, and 0-9.

2:

Make sure that the lines in the example screenshots for the game on vndb are in the script.

Sometimes these lines might have been edited in a patch, so don't treat a missing line as a definite error.

3:

If the VN ever uses ruby text, find an example of ruby text inside the VN, then look for it in the script. Check for other variations on the ruby text formatting. Make sure that your ruby text formatting replacement works well with all the ruby text syntax the engine uses.

4:

Search the script for the names of some of the characters. You might find places where they're part of a speaker name or a filename instead of content text.

5:

Search for half-width and full-width spaces. You should never see spaces in the middle of words or set phrases like "飛　行機" or "いいだ　ぜ", but they can appear when removing bad forced line breaks. Having spaces in situations like these makes the analysis tools get confused.

6:

Look at the beginnings of several scenes. Make sure they don't start with obvious scene or room names or anything.