Nature excels at combining dissimilar materials, endowing biological composites with mutually exclusive properties and enabling complex functions. For example, the armored skin of the porcupinefish (Diodon holocanthus), a long-spined pufferfish, consists of an array of highly mineralized, rigid spines embedded within a highly stretchable soft skin [1]. Inflation of the body is accompanied by the erection of the spines, providing an effective defense against marine predators [2]. In this work, we employ three-dimensional nonlinear finite element simulations to investigate the mechanisms governing spine rotation in a parametric pufferfish skin model, comprising a periodic hyperelastic soft-skin unit cell with stiff spines arranged in a triangular lattice and subjected to axial stretch. Simulation results demonstrate that spine rotation is not governed solely by spine arrangement but rather emerges from the complex skin–spine interfacial morphology. Two structural features are identified as necessary (but individually insufficient) mechanical conditions for the onset of interface-driven rotational instability: (i) a sub-needle cavity that introduces controlled mechanical decoupling at the spine base and establishes a kinematically admissible displacement pathway for the initiation and amplification of rotational motion; and (ii) a collagen-rich, tendon-like interfacial region that serves as an effective force-transmission zone between the compliant skin and the stiff spine. This behavior cannot be captured by fully bonded spine–skin structures. Quantitative simulations reveal that significant spine rotation—reaching approximately 60° under finite skin strains within experimentally observed ranges—occurs only when sub-needle release and tendon-like interfacial stiffness jointly satisfy the mechanical criteria governing interface-mediated rotational instability. By reframing pufferfish spine rotation as a nonlinear interface mechanics problem, this study provides the first mechanism-based quantitative description of the phenomenon and establishes a robust computational foundation for the design of next-generation bio-inspired soft robotic systems with actively reconfigurable architectures.