Hi maintainers, would a small Chinese moderation example be useful?

I have a working prototype with 12 synthetic cases covering abuse, quoted reports, and prevention messages. It requires no API key and uses fixed predictions to demonstrate false positives, false negatives, precision, recall, and F1 through existing Promptfoo features.

I saw eval-f-score and eval-moderation. Would you prefer a new example or an addition to one of those?

The prototype was tested on Promptfoo 0.124.1. The cases and proposed labels were AI-assisted and still need human review. I'd welcome your guidance before preparing a PR.