We often think that intentions play a crucial role in explaining the im/permissibility of actions. (Compare 'collateral damage' in strategic bombing vs. intentionally targeting the population. Or buying your neighbour rat poison to help him kill rats vs. to help him kill his wife.) But Scanlon offers a nice general strategy to undermine this position: (1) Consider the impermissible action. (2) Cut out the bad intentions, and suppose the agent is instead simply acting negligently. (3) Notice that the result is still impermissible.
For example, if you have good grounds for believing that your neighbour is trying to kill his wife, it's presumably impermissible to buy him some poison, no matter how empty (of bad intentions or good sense alike) your head is at the time. What matters to the question of permissibility is the expected consequences of the action, and your private intentions make no difference here. We are tempted to think intentions matter in the original cases because they covary with other factors that matter, e.g. what you can reasonably expect your neighbour to do with the poison. But once we separate them out (by appealing to cases of negligence) we see that it's these other factors, and not one's intentions, that typically matter for questions of permissibility.
To reinforce this conclusion, notice how bizarre it would be if perverse motivations could typically render otherwise permissible actions impermissible. Scanlon discusses the case of a doctor justifiably administering painkillers that will foreseeably end the life of a terminally ill patient. Supposing this is permissible in the ordinary case, does it change when the doctor happens to delight in having an excuse to kill his patients? Should he say, "Sorry, I can't deliver the lethal painkillers, because I know that if I did so I would intend not just the cessation of your pain, but also your subsequent death. Not that it would change anything, but you know, it's impermissible to intentionally kill people. Good luck finding a doctor who intends only the former of the two predictable consequences of the procedure!"
This is a second general strategy that can serve to undermine purported links between intention and permissibility: just combine the two contrast cases into one where the agent may have either intention in performing the action, and note that it would be completely bizarre for this to affect its permissibility. In the bombing case, for example, suppose a single potential missile target could serve both military and terrorist objectives. In deliberating about whether to press the launch button, must the general first introspect on her own motives? If she finds that she intends the civilian casualties, should she summon her more pure-of-heart lieutenant to press the button instead? Who could possibly care?
What matters for permissibility are the reasons for action that exist in the world. Once we establish that there is sufficient reason to bomb the target, we've established that the action is permissible. The actor's intentions don't change the moral status of the action. What they should influence is our assessment of the agent. But that is to raise questions of blameworthiness, not permissibility.