While I personally would not pretend to know anything about it, I would think logic would dictate that if one had anything you wanted to secure behind encryption if you had the ability you would want to produce your own. Anything “off the shelf” would mean you could never be certain that what you were trying to secure, actually was secure.
As for me, personally, the answer is simple, do not have behavior or discussions that require such security. The old expression, “The only way two people can keep a secret is if one of them is dead.” comes to mind.
It is not impossible that the NSA could have fooled the world's experts and gotten positive reviews of algorithms (e.g. AES) with security flaws. But if that is the case it would be because of a vastly superior and completely secret NSA technology to brute force or bypass the security. Brute force is plausible and described in academic literature but with technology that is completely speculative (e.g. large scale quantum computing). Bypasses are not in the literature AFAIK and would be unlikely with the exception of minor weaknesses that would reduce the brute force computing requirements. See Schneier for descriptions of those.
There is also evidence that the Chinese have rolled their own perhaps for those reasons. But that is probably a bad idea given alternatives like adding rounds to a public algorithm. If you simply add rounds you greatly increase the brute force computation requirements at the cost of a bit more computing on your end.