That score idea from yesterday was too simple.
I need more than that.
I'll have a reports object in the context. Each report has a name, a score, and an explanation. The script can add a report anytime. Can ask for a summarized score. Can list them.
This lets me run ryoms in "explain" mode, where it will format the reports in the end, instead of acting like a milter. Will still stop at the same time.
One thing I can add ahead of time is the scoring system. I can make score available in the context, an f32 that every callback can fiddle with as they please.
The decision what to do with the score is also the responsibility of the script. The runtime merely provides a way to track it.