Camouflaged Object Detection (COD) has become increasingly important in multimedia systems, aiming to enhance visual understanding and accuracy. Traditional methods often treat all camouflaged objects uniformly, ignoring their unique characteristics such as occlusion and undefined boundaries. To address this, we propose a novel method called Attribute Classification Guided Transformer (ACGFormer). ACGFormer features two key components: the Attribute Classification Guided Branch (ACGB), which uses object attributes to guide detection, and the Object Segmentation Branch (OSB), which employs a Detail and Texture Extractor and Cross Agent Attention modules to capture and refine camouflaged details. The OSB also incorporates a Feature Adaptive Refinement (FAR) strategy for improved prediction accuracy. Our extensive experiments on various COD datasets show that ACGFormer achieves state-of-the-art performance, surpassing previous methods significantly. Code will be released at https://github.com/Lwt-diamond/ACGFormer .