Spec-kit for a complex brownfield project

[ranp]

I am looking for guidelines or the community experience regarding using spec-kit for a complex feature-rich brownfield system. There are multiple challenges I seek advice on:

1. We already have architectural guideline documents, as well as programming how-tos in the form of .cursor/rules/ markdown files. Can these be just referenced from constitution.md or we must somehow bake them into the file?

2. Our codebase contains many repos, and a typical feature requires changes in multiple ones (for example, the web app repo, some microservice in a separate repo, and some common module used by the microservice, in a third repo). In Cursor, we open multiple repos into the same workspace, and the agent works well for narrow tasks focus on a specific component, that need a wider context (e.g. an API they can consume) . But a full end-to-end feature seems different. Are there best practices as to where Spec-Kit directories should reside, how to tell it where to put which code part, and whether it handles this use case well in general?

3. Following the Methodology document, when some aspect of an existing complex feature changes, we should change the feature’s spec. However, re-implementing the entire feature is unrealistic. Therefore, I see a separation between the spec as source-of-truth, and the feature change effort. To plan and define the change tasks, the agent must know the Spec before and after the change. Are there clear guidelines on how to approach this? ( I found discussion Evolving specs #152 highly relevant, but the conclusion is not clear to me)