AGENTS.md contains your preferences for how you want your agent to behave.
Think of it as the prompt your agent reads every time it starts.
Mine tells the agent to always read the project's README and CONTRIBUTING files, so it has the context of the repo I'm working on.
Skills on the other hand are repeatable flows you can reuse over and over again.
Architecture is a great candidate for a skill. My compose-architecture skill describes how my app's screens should be built.