oqoqo-inc/oqoqo

Build a complete eval set in Oqoqo from a product's documentation, including the tasks, the machine, the files, the assets and the treatments. Use when someone points at a product and asks for an eval set to be created, or asks what they should evaluate about their product. Covers what to ask them, how to find the product's MCP server, CLI and SDK, and how to write everything into an Oqoqo project.

Build the machine, fixtures, treatments and seeded data an eval batch runs on, so that parallel runs stay valid and the baseline stays clean. Use when setting up an experiment environment, deciding what belongs on a machine versus a treatment, or seeding a product account with test data.

Turn a product's own documentation into a batch of agent eval tasks with fair rubrics. Use when someone points at a product's docs, README, or skills repo and wants tasks that measure whether an agent can actually use that product. Also use when reviewing or fixing an existing rubric.