For context.md, it's that simple because it's unstructured data extraction from your codebase and working with a regular LLM model.
For design.md, I have a prompt because we're now working with a reasoning model and doing structured data extraction: create me an issue on Linear with a title, description, and a to-do list.
I would recommend trying the approach yourself and saving the prompts if you can nail down the repetitive asks.
