This is a double doubt at the Torah level. The first doubt is whether twilight has begun. The second doubt is that even if twilight has begun, there is still doubt whether it is day or night.

That cannot be right, because if so the whole discussion about rabbinic prohibitions during twilight would be unnecessary. This would be a rabbinic doubt, where we rule leniently. We should discuss why that is not correct, and the simple explanation is that the Sages established that twilight has the status of certainty and not of doubt (were it not for the rule regarding rabbinic prohibitions during twilight). And as stated, there is also the law of adding onto the Sabbath and the community’s acceptance of the Sabbath.

But from the Mishnah in tractate Shabbat it seems that twilight is a doubt and not a certainty:
“If it is doubtful whether it has grown dark and doubtful whether it has not grown dark, one may not tithe definite produce, and one may not immerse vessels, and one may not light the lamps. But one may tithe demai, and prepare an eruv, and insulate hot food.”

And in Maimonides’ commentary on the Mishnah it says this:
“And we explained in the second chapter of tractate Terumot that it is forbidden to separate terumot and tithes, and likewise to immerse vessels, on the Sabbath, and this will be further explained, along with its reason, in tractate Beitzah. Therefore it is forbidden to do so during twilight because it is a doubt, and this is what they refer to here as: doubtful whether it has grown dark and doubtful whether it has not grown dark.”

Maybe I wasn’t clear enough. Of course twilight is a doubt. That is explicit in many, many places. But the Sages were stringent about it in some matters (such as rabbinic prohibitions during twilight), treating it as certainty. Regarding rabbinic prohibitions, for example, it is clear that they treated it as certainty, because otherwise we could be lenient with any rabbinic prohibition regardless of the dispute between Rabbi Yehuda HaNasi and the Rabbis, and even not only for the sake of a commandment, a major need, or preventing a loss. And if regarding rabbinic prohibitions they treated it as certainty, it is reasonable that the same applies to Torah prohibitions of labor. The leniencies in cases of doubt during twilight are not with prohibitions of labor.
However, the Magen Avraham raised the possibility of following the presumption, and then there would be a difference between twilight at the entrance of the day and twilight at the exit of the day (the presumption of weekday or the presumption of Sabbath). But this is a lone opinion, and it is very difficult in light of many sources.

Regarding your doubt whether the unusual manner has to be in something that is part of the essence of the act:
1. See Rabbi Shmuel Ariel’s article in Techumin 31
2. See Shemirat Shabbat KeHilchatah in its introduction to the laws of the Sabbath, where he cites in the name of Maharach Or Zarua that placing a pot on the fire with one’s elbows is not considered an unusual manner, and what he discusses regarding his words (Rabbi Ariel also addressed this in the article mentioned above)

I thought again about this question: maybe what the Sages were stringent about regarding twilight, treating it as certainty, applies only to things that are not in the category of a commandment, a major need, or loss, but regarding a commandment, a major need, or loss, they were not stringent about twilight to treat it as certainty. If so, a Torah-level doubt during twilight would count as a double doubt at the Torah level, which is ruled leniently. Right?

Something here is off. I can’t manage to correct and delete my previous answer. It is definitely possible.

Doesn’t this depend on the question whether we say that a double doubt can be reversed, and in practice one is stringent? No?

I don’t understand the connection.

After all, the double doubt is whether twilight has begun, and even if it has begun, maybe it is still day and not night. There is no way to formulate the double doubt in the reverse way: maybe it is day, and even if it is night, maybe twilight has not begun. If so, it turns out that this is a double doubt of one type, in which, as far as I recall, the practice is to be stringent.

Exactly. Just like the case of doubt whether a lion entered and doubt whether it mauled. That really is a difficult question as to why this is considered a double doubt at all, even aside from reversibility. There it is a single doubt.