<p>This paper presents a robotic fabrication framework that enables skilled carpenters to communicate fabrication intent through hand-drawn markings rather than relying on code or screen-based interfaces. Inspired by the Japanese sumizuke tradition, where ink markings guide assembly and communicate spatial relationships among carpenters, the system translates visual cues into robotic actions for timber construction. The workflow integrates a distributed software architecture combining computer vision, semantic interpretation, and robotic control. A deep-learning-based keypoint detection model identifies custom symbols on timber surfaces and maps their geometry through a modular software architecture. The framework was tested across two robotic setups: an industrial KUKA KR10 and a collaborative UR10e, through experiments involving absolute and relative localization, marking-defined features, and on-site joinery fabrication. Results demonstrate that markings can serve as a robust, low-threshold interface for human–robot co-fabrication. The proposed workflow achieved millimetric precision with robotic touch-up and approximately ± 3&#xa0;mm accuracy from vision alone. This study establishes a craft-native, CAD-independent approach to robotic fabrication that adapts to the logic of carpentry workflows and bridges traditional craftsmanship with contemporary automation. By reducing the technical and financial overheads associated with conventional CNC systems, the framework opens new perspectives for adoption among SMEs, where renovation, retrofitting, and material reuse are core to the daily practice.</p>