What problem does it solve? Refactoring or adding modules to a topological data analysis (TDA) pipeline often produces shallow adapters, duplicated logic across battery scripts, and unclear module boundaries that silently alter committed research results. This Skill provides a disciplined procedure for defining module boundaries and interfaces around clean scientific seams. ## Core Features & Use Cases - Seam Vocabulary: Names modules against a fixed vocabulary (trajectory_loader, embedding_transform, null_generator, diagram_computer, permutation_test, result_writer, and more) so boundaries stay consistent. - Full Interface Definition: Forces interfaces to cover invariants, error modes, seed handling, performance, provenance fields, and output schemas — not just function signatures. - Paper-Result Risk Control: Flags any refactor that could change committed result values, requiring a rerun plan and a [DECISION] record instead of a silent change. - Use Case: When adding a new null-generation step to a persistent homology pipeline, use this Skill to decide whether it is its own seam, define its contract surface, and plan migration of existing result JSONs. ## Quick Start Ask the AI to apply the tda-codebase-design skill to plan the module boundary and interface for a new permutation testing step in the TDL pipeline.
