When I write and execute a prompt in Cursor IDE, what specific information is included in the context that gets sent to the underlying language model (such as GPT or Claude)? E.g., does it include previous messages, file contents, or other contextual data from the IDE?