Well, if you thought my points 1 and 2 were your points, then I guess we're just not communicating very well!
We can generalize about deeds without knowing intent because many deeds we consider good it's hard to do for a wrong reason, and many deeds we consider evil it's hard to do for a right reason (especially when there is an absolute law against them--a rational person not under duress cannot commit adultery without sinning).
So throwing out the whacked-out situations in which God told people to commit infanticide, in a normal situation we are ruled by the law against murder. Therefore we would obviously intervene if we saw someone attempt to kill a baby--there is no circumstance under which that could be a right action, although if the person is retarded, insane, or senile they might not be guilty of attempted murder.
Breaking it down to the very simplest level, I say that physical deeds are not morally significant because it is the mind behind them that is responsible. If you built an android that looked exactly like a human, it could commit all of the physical deeds we think of as evil or good, yet it would be incapable of either pleasing or displeasing God because it has no moral sense.
So the android donates money to charity, helps little old ladies across streets, and memorizes the entire Bible in half an hour. Then it has a short-circuit, knocks someone down for simply being in the way, steals a car, and breaks the speed limit. So what? Its actions are amoral.
You might as well say a crocodile sins when it drowns a swimmer.
If at some point in the future such androids become common, one that kills someone will be feared and destroyed, but it will not be blamed for having sinned. If we can call some action it does evil, it is evil only if the android has been programmed to do that deed by a moral agent with evil intent--a person who is responsible for the android's actions.