We all carry SSH keys around in our agents like they are digital pocket change. It feels convenient, but if your machine gets compromised, that agent is an open vault for an attacker to pivot through your entire infrastructure. I learned this the hard way after a developer had their laptop pinched at a conference and we spent the next six hours frantically revoking access across a dozen jump boxes.

Why agent forwarding is a liability

When you use ssh -A, you are literally piping your local socket into the remote host. Any user with root access on that server—or an attacker who has gained a foothold—can simply look at your environment variables and point their own session to your forwarded socket. They don’t even need your private key; they just need to impersonate your connection while it is live. It is essentially giving away your identity to every host you touch.

ssh -o ProxyCommand='nc -X 5 -x 127.0.0.1:9050 %h %p' user@internal-host

• Stop using ForwardAgent in your ~/.ssh/config files immediately
• Use ProxyJump to tunnel traffic instead of giving the remote host a socket
• Require FIDO2 or hardware-backed keys that force a physical touch for every signature

The alternative is ProxyJump. It moves the complexity to your local side. The server never sees your agent; it just handles the encrypted TCP stream passed through a legitimate gateway. It is slightly more verbose to type, but it stops the risk of credential leakage dead in its tracks. If you are still relying on legacy agent forwarding, make the switch today before you find yourself auditing every single log file on your network during an incident response exercise.