No, I don't think that's how it works. A class break in a core cryptography primitive or even a major break in a particular crypto format would be one of the most closely protected SIGINT secrets in the country; the number of people who would be exposed to even knowledge of its existence would be very low.

Agee was higher up in the intelligence hierarchy than Snowden and the MINERVA secret, while a fairly big deal, is not nearly as big of a deal as 'NSA can break some kinds of modern crypto' would be.

More importantly, I think you're misreading what the new writeups on this story say about Agee's knowledge. He doesn't mention MINERVA and didn't know anything about the BND-CIA joint infiltration of the company. Here's what he writes in Inside the Company:

The National Security Agency cannot break this code system mathematically but they can do so if sensitive recordings can be obtained of the vibrations of the encrypting machine when the discs clack to a stop. The recordings are processed through an oscilloscope and other machines which reveal the disc settings. Knowing the settings, NSA can put the encoded messages, which are intercepted through the commercial companies, into their own identical machines with identical settings, and the clear text message comes out. Although the Swiss manufacturer when selling the machine emphasizes the need to use it inside a sound-proof room on a table isolated by foam rubber, we hope this particular code clerk is careless. If we can discover the settings on this machine in Montevideo, NSA will be able to read the encrypted UAR messages on the entire circuit to which their Montevideo Embassy pertains.

(The speculation is, of course, with reasonable circumstantial support, is that it is a ~$1B program that has brute-forced the most common 1024 DH group in use.)

We simply don’t have the hard data, it is (educated) speculation based on what information we do have.

Even the existence of the program is TS. Its capabilities remain secret, due to the exact system you describe.

"No decrypt available for _this_ PGP encrypted message."

You don't write an error message that way unless the code has a success case as well.

(Unrelated: As well as TLS long term keys, passwords, hashes, usernames, and any other kind of metadata or secrets that may be useful one day in the future, if nothing more than for dictionary attack prefix/suffix fodder.)

I wouldn’t be surprised if they have some more creative secret key sources too: stolen and glitched smart cards, laptops that disappeared out of targets’ cars, tossed offices via evil maid, dumpster diving, all of it, including some I probably haven’t thought of because I’m a computer nerd and not a military intelligence cloak-and-dagger type.

Put all the recovered secrets into a big ol’ database, because disk is cheap and keys are small. Keep it for all time, Just In Case.

Of course, there is a request system frontending this capability.

There are many PGP messages they can decrypt, simply because they slurped up the specific private keys for those messages at some point, and simply saved every secret key, hash, or password that they ever saw, as a general organizational policy.

That doesn’t mean they have broken PGP.

I'm sure he said "to my knowledge" or something to that effect. That is, at least for at least relatively far into the circles of confidence, people did not know about encryption being broken algorithmicly or PGP broken in practice.

Russia has world class cryptographers too, and may have beaten the NSA to the punch. Snowden is after all currently living under FSB protection, which I doubt came for free. Someone willing to sell out their country would likely sell out its people too.