def score_case(case: dict, result: dict) -> tuple[bool, str]:
    answer = str(result.get("answer", ""))
    expected = str(case.get("expected_answer_contains", ""))
    if expected and expected.lower() not in answer.lower():
        return False, "answer substring mismatch"
    if len(result.get("tool_calls", [])) != int(case.get("expected_tool_calls", 0)):
        return False, "tool call count mismatch"
    return True, "ok"