An attack is demonstrated on a previously proposed class of key agreement protocols. Analysis of the attack reveals that a small change in the construction of the protocols is suucient to prevent the attack. The insight gained allows a generalisation of the class to a new design for conference key agreement protocols.