Why are "how to" questions requiring knowledge beyond a particular implementation, typical of questions at the planning stage, considered not useful?
Consider the question "How to use Symfony's AssetMapper with a template engine in general, not specifically Twig?" The question is asked because all tutorials, including those of the official AssetMapper documentation, explain AssetMapper in terms of Twig. A similar question could be closed because it asks for software recommendation. But no software recommendation is explicitly asked here: any knowledge about how to use AssetMapper in general (in a way that is not tightly connected with Twig) will answer the question. It is only because an example is very useful, almost required in practice, that a software beyond Twig is indirectly asked in the question.
This is very different from question Is asking the community to find (not recommend) an off-site resource allowed?, because it considers a case where a request for off-site resource is only indirect and the question goes way beyond that.
Note that posting a question that is not connected to a specific implementation usually means that the asker is at a planning stage. It is not a coincidence that the question "How to use Symfony's AssetMapper ... not specifically [with] Twig?" emerged in such a context. Conversely, questions at a planning stage are not connected to a specific implementation, almost by definition. So, "planning stage" and "not connected to a specific implementation" are different ways to formulate the same issue.
The issue is more easily connected to software recommendation when we express it in terms of a lack of specific implementation. To answer a question that is not about a specific implementation it is useful, almost necessary, to consider different software or off-site resources that give examples, for example, to compensate for a lack of a specific concrete implementation. So the question will easily be interpreted as asking about a software or off-site recommendation. This is what happened with the question about AssetMapper.
My theory is that if you ask a question that arises along a path that is taken by many it is likely that someone would have met the same question and can answer it. Also, the question will be useful to all those who will eventually take the same path. This is the case for questions that arise in the context of a curriculum that is followed by many schools or universities or in the context of developments done by big enterprises. Questions that arise at a planning stage by an individual are not of this kind, because it is as if the asker is still in a path among many other possible paths and has not yet taken a highway used by many. This distinction exists at all level of complexity.
What is disliked here is that the question is difficult to answer, though it can be a clear question that is answerable. This is also the reason, the asker is asked to explain what he has done already in their implementation. This is because there is a hope that it will bring the question into a highway taken by many, but it is a misunderstanding that, fundamentally, the question is where it lies, not in a highway. Attempts to bring the question into a highway will only bring confusion. The criticisms will often be that the asker did not do enough work. They will be compared to a student that did not do their homework.
This makes me sad, because questions at the planning stage can be answered in a very precise manner by those who have the knowledge and these answers would be very useful and the "no off-site resource" criterion does not directly apply. It is an abuse of the criterion.