Normalizes a surviving-mutant record across StrykerJS, PIT, mutmut, and Mull into one shape, classifies why it survived (missing case, weak assertion, equivalent mutant, unreachable code, flaky killer), applies per-mutator heuristics for conditional-boundary, arithmetic-operator, statement-removal, and constant mutations, and drafts the specific test that would kill it. Includes the full read-only investigation workflow: take a mutation report (Stryker JSON / PIT XML / mutmut output / Mull JSON) plus the repo at the same commit, read each survivor's mutated line and covering tests, classify, and propose - never auto-rewriting tests. Treats equivalence as a judgment call, because deciding whether a mutant is equivalent to the original is undecidable in general, so a residual survivor rate is expected rather than a defect. Use when a mutation run has finished and the report lists surviving mutants (typically 5+) that nobody has yet explained or turned into concrete test cases.