moronic monday (pst) question time. this may have been covered elsewhere, and emphatically shot down (and if so, I apologize), but I'm interested in using ssh-agent to decrypt data. ssh certificates are the most secure user auth token we have, and even though ssh certs aren't chainable, we routinely use them to assert identity to things other than ssh (eg. if you can present a cert that's signed by a known ca, and you can prove you have the private key that belongs to the cert, we know who you are). So it'd be nice to be able to give some data back to the client, encrypted with the client's pubkey, that the client could then ask the ssh-agent to decrypt using the corresponding private key. in the past, we've generated a new keypair and sent the new pubkey with the request, but that's a little clunky and it's as tightly bound to "one source of truth" as if we could just rely on the key from the cert. I'm guessing there's a dead simple reason something like this isn't supported already. like, maintenance of code that only a weenie like me would use or that it would enable some new class of cryto attacks or something. Cheers, peter _______________________________________________ openssh-unix-dev mailing list openssh-unix-dev@xxxxxxxxxxx https://lists.mindrot.org/mailman/listinfo/openssh-unix-dev