概要
The Propagators skill enables Claude to utilize the Sussman/Radul model of propagator networks, a sophisticated alternative to traditional dataflow. It allows for monotonic, concurrent, and bidirectional information flow, where cells in a network gain information until a stable fixpoint is reached. This is particularly useful for solving interdependent constraints in CAD modeling, physical simulations, and advanced type inference where information must flow from both inputs to outputs and vice versa, while providing robust dependency-directed backtracking to resolve logical contradictions.
