I agree with Draconis's answer that a future less vivid conditional is the best choice here, but I think the specific subtype of future less vivid that's most appropriate is the type with a perfect subjunctive in the protasis, rather than a present subjunctive.
In both types of future conditions (future more vivid and future less vivid), the protasis can use a perfect tense to show that the action described in the protasis is complete before the action described in the apodosis, rather than being simultaneous with it. In a future more vivid, this yields a future perfect indicative, while in a future less vivid it yields a perfect subjunctive.
Here is Allen and Greenough's discussion (516c):
If the conditional act is regarded as completed before that of the apodosis begins, the Future Perfect is substituted for the Future Indicative in protasis, and the Perfect Subjunctive for the Present Subjunctive:—
sīn cum potuerō nōn vēnerō, tum erit inimīcus (Att. 9.2A. 2), but if I do not come when I can, he will be unfriendly.
sī ā corōnā relictus sim, nōn queam dīcere (Brut. 192) , if I should be deserted by the circle of listeners, I should not be able to speak.
This use of the perfect doesn't seem to be a hard-and-fast rule but more of a tendency or authorial choice, but it seems apt in your case, since Alice would first fall of the top of the house and only then say or not say anything about it. So I would use ceciderim in the protasis and dicam in the apodosis (which also makes it clear that this is a future less vivid condition since it gets around the issue Draconis mentions of cadam, dicam both being identical with future indicatives).