This follows Sample size calculation from two ROC curves with a different ratio of cases to controls than sample data on Stack Overflow.

In power.roc.curve.roc, kappa is silently ignored if two roc curves are passed. It is always calculated from the data itself.
In addition if a single ROC curve is given, kappa is passed down to power.roc.test.numeric, but defaults to 1, instead of the ratio in the data, which is somewhat inconsistent.

Aim:

• Kappa can be changed by the user for 2 roc curve. Defaults to ratio in data.
• Same behavior for 1 roc curve

Additional note:

• Check if tests already exist for kappa of single ROC curve, or add them if necessary.