[msavchen-nasa]

No description provided.

[scastro-nasa]

is this needed because the "forward" transform was already looked up and being used by a previous behavior? Because otherwise I might recommend just calling the "lookup transform" behavior but with the arguments reversed?

[ndunkelb-nasa]

I think this looks fine after the minor comments! Would be good to add in a test for some simple inversion cases too!

[msavchen-nasa]

is this needed because the "forward" transform was already looked up and being used by a previous behavior? Because otherwise I might recommend just calling the "lookup transform" behavior but with the arguments reversed?

I tried doing that,
the problem with that is that in order to get b_T_a we have to publish a_T_b to the tf buffer first.
With static transforms publications the buffer can keep both of these transforms and create something like this on the tf graph

graph LR
    A((a)) --> B((b))
    B --> A 

Loading

this behavior should let you avoid that.