The paper proposes a generic approach to assembly planning where individual tasks, with detailed technological content specified by features, must be combined into an optimal assembly plan subject to technological and geometric constraints. To cope with the complexity and variety of the constraints that refer to the overall assembly process, Benders decomposition is applied. The macro-level master problem looks for the optimal sequencing and resource assignment of the tasks, while sub-problem modules ensure plan feasibility on the micro-level from aspects of technology, fixturing, tooling, and collision. Constraints are also dynamically generated for the master problem. The approach is demonstrated in automotive assembly.