I'm not a lawyer either but the general guideline would be - no deal names, no customer lists, no identifiable information, no pricings, no description of strategies. Speak in general terms and talk about tackling issues that only someone that has modeled would identify.
When it comes to transactions or client names I also ensure that the information isn't easy to reverse engineer. e.g. I once covered commodities and power and when referring to a particular problem on a project we resolved I was always vague on the country it was in as it was the only power plant within the region of that country it was in and often clients would have worked on the same transactions as I did.
In terms of the modelling I would try and write down what I did and see does it read as if it tells firms anything about strategy and if they could rebuild a model. Again I speak in general terms. I'd err on the side of caution even when talking about non-proprietary models as you could wind up giving away which publically available model your firm uses. But the issues you encounter tend to be universal e.g. I once priced CDOs using a Gaussian copula model. The model is well known and tons of papers show you how to use it. Even then in interviews I refer to "a pricing method" and some of the issues I encountered in getting efficiency and accuracy from it - then it would have been easy to see I was not bluffing.
Again, I'm not a lawyer but maybe tap up a friend that is a derivatives lawyer to be sure?
Also be aware some firms are unscrupulous about this. One trader friend hadn't exactly had the best experience at his first firm and was trying to move - several times hedge funds took the approach of "well it appears you haven't been trained properly" (which to be fair was true) and going "have you any examples of research you did with them?". I'm sure you know how to handle this but just wanted to be aware it has happened to some.