testdata-quality

# Test Data Quality Skill Final test data must pass: 1. `problem_verify_tests` checks: `file_count` / `answer_consistency` / `validator` / `no_empty`. 2. `limit_ratio` and `limit_semantics` (type=3 and type=4 must not be semantically overlapping). 3. `wrong_solution_kill`: every wrong solution must be killed by at least one test. ## Additional Required Checks - Whether test points contain obvious duplicates (low information gain). - Whether type=3 and type=4 are semantically distinct rather than simple parameter scaling. - Whether pass/fail behavior of reference/wrong solutions matches expected roles. ## Output Template - `decision`: `go` / `no_go` - `failed_checks`: list of failed checks (with reasons) - `coverage_report`: scale distribution, limit-case ratio, and wrong-solution kill statistics - `repair_priority`: fix priority (`P0` / `P1` / `P2`) - `reverify_plan`: re-validation order after fixes ## Decision Rules - `go`: all required checks pass, limit semantics are valid, and wrong-solution kill is effective. - `no_go`: any required check fails or coverage quality is insufficient. ## Forbidden Behavior - Do not approve final tests if any required check fails. - Do not accept type=3 and type=4 when semantics are effectively identical. - Do not skip wrong-solution-kill validation when wrong solutions are available.