Feature request for product/service
– Other –
Describe the request
Please keep the tasks and test sets private, but it would be helpful to your customers if we had a solid Cursor-based evals harness to use.
For example, our company has many rules, skills, and MCP servers across two private plugins. They help us unify our workflow across the entire team. From end to end, Composer (successfully) completes tasks in minutes while Grok takes hours. Narrowing down the problem has many dimensions to it, so a proper evals harness would help tremendously.
Operating System (if it applies)
MacOS