Design Pattern: using `rlang::eval_tidy()` in a secure way.

[fb-elong]

Is rlang::eval_tidy() evaluating arbitrary input expressions? That would create a security vulnerbility, as the expression can be malicious.

[nanxstats]

I did some preliminary research and wrote a draft.

The basics seem clear but I wonder if there is a more authoritative guide on this topic.

[nanxstats]

We can close this for now. We can analyze the use cases on case by case basis. It's a bit challenging to write a comprehensive secure evaluation guide from a non-rlang author's perspective.