I've redacted the MD5s in the above posts.
I've also clarified my opening post. But basically, common sense should prevail: you shouldn't publish the hash for a password you want to be secure.
To the best of my knowledge, MD5 hashes are not currently reversible - meaning it's not (currently) computationally feasible to find an input that produces any given MD5 hash.
But there are other known flaws in MD5, such as the ability to produce two files that hash to the same value - although you can't control what that value actually is.