Canon DSL v2.1 is a metadata-first YAML schema for distilling solved mathematics problems into structured, machine-readable records that can support synthetic problem generation. The paper describes a practical workflow in which a solved source problem is first converted into a rich metadata record containing objects, givens, asks, parameter domains, invariants, theorem roles, answer normalization, solved-instance snapshots, and computational checks. That metadata can then be used to forge and mutate same-family descendants while preserving lineage, route structure, and answer contracts. The work presents the Canon DSL v2.1 schema, the distillation procedure, prompt-level heuristics, mutation operators, a checking contract, and a three-generation number-theory case study. The aim is to make synthetic mathematical data easier to generate, inspect, package, and reproduce from a single structured source.
Aaditya Paudel (Wed,) studied this question.