NoEskape

well, yeah, there is some nuance to this.

you can "train" your --p in such a way that even for neutral prompt, nijijourney will be gravitating towards the style defined by --p. That requires you to pick the most questionable images, when training --p (skipping a lot of image pairs that are too neutral)

you can find existing --p (or interesting prompts that worked) by looking for questionable images in Explore gallery(at nijijourney.com), looking for non-yet-banned questionable search terms like "bustier" , "body", "stocking", "stockings", "leg", "thigh", etc. Make sure to use "--no anime" when using found prompts and/or --p , to avoid anime looks

Another approach (that can combined with above) - is to use suggestive words in prompts, that are not banned yet. "sexy woman" can go a long way with various --p and --sref. Keep in mind that certain questionable words are banned "in context". For example "cinderella as lingerie model" prompt works, but can be blocked for other characters, or if you add any other words into the prompt (though it depends).

Just avoid running into several AI moderation messages in short amount of time (2 or 3 in a row can be a problem eventually). you will either ger auto-banned or will be manually reviewed (and manually banned). Sprinkle failed attempts with neutral prompts