The paper proposes RAGC, a novel robust attributed graph clustering method that combines hybrid-collaborative augmentation and contrastive sample adaptive-differential awareness to obtain more discriminative representations.