What problem does it solve? Building a competitive programming test suite that is legal, reaches every declared bound, and separates subtasks by complexity is error-prone when done by hand. This Skill produces the full test-data contract — checker, validator, and generators — using testlib, so wrong or slower solutions fail and the intended solution passes. ## Core Features & Use Cases - Checker and Validator Authoring: Writes testlib-based checkers (stock like ncmp/wcmp/rcmp6 or custom with registerTestlibCmd) and validators driven by a generated constraints.h header, with per-subtask group branching. - Generator Families: Produces random, max-size, boundary, structured-adversarial, and hand-written generators, with test counts split by subtask point weights and ICPC/OI kill policies. - Subtask Separation Enforcement: Verifies empirically that weaker per-subtask solutions score zero on stronger groups and that every declared bound is attained by some test. - Use Case: Given a problem with subtasks g1 (N ≤ 1000, O(N²)) and g2 (N ≤ 10⁶, O(N)), generate a validator that enforces both bounds and a 50-file test suite where g2 tests reach N = 10⁶ so the O(N²) solution cannot leak points. ## Quick Start Ask the assistant to write the testlib validator, checker, and generators for the problem in the current directory and build the subtask-grouped test data.
