<p>Continuous carbon fiber reinforced plastic (CCFRP) components offer exceptional specific strength and stiffness, yet their additive manufacturing (AM) for complex geometries remains a challenge due to the limitations of traditional planar slicing and support structure requirements. This paper proposes a feature-based adaptive slicing and path planning algorithm tailored for robotic CCFRP additive manufacturing. The proposed approach first identifies explicit and implicit feature boundaries of STL models to implement an automated model decomposition strategy, partitioning complex geometries into simple sub-volumes. Subsequently, a multi-directional adaptive slicing algorithm is developed based on feature normal to generate curved layers, effectively eliminating the need for support structures while enhancing interlaminar performance. To ensure fiber continuity, a smooth spiral toolpath generation method using geodesic-based offset algorithms is introduced. The effectiveness of the algorithm was validated through the fabrication of a complex rotor model. Experimental results demonstrate that the proposed method achieves high dimensional accuracy with a mean relative error of only 1.94%. Furthermore, compared to conventional planar slicing, the feature-driven adaptive strategy significantly improves the interlaminar bonding and radial tensile strength of the components. This research provides a feasible solution for the high-performance robotic manufacturing of complex CCFRP structures.</p>