Screen text and files with typed questions before reading them into agent context.
Use Jev to decide which files, webpages, tool descriptions, messages, or search results deserve the agent's attention. Screen batches against a query without loading full file or page contents into the main context first.
