Term · Tracking setup
Attribute
Definition
A capability your prompts test, grouping many differently-worded prompts under one heading, such as file management or on-device privacy.
How it is calculated
Attributes exist because a single prompt measures its own wording rather than the capability it was written for: swapping one label can move visibility by tens of points, so one prompt per attribute answers a question about phrasing. Two or more differently-worded prompts under the same attribute turn that into a finding. Prompts with no attribute still run and are counted; the attribute report states how many are ungrouped rather than leaving them out of view.
Where to see it
This term describes part of how refd collects and scores AI answers. The methodology page sets out the full contract, and the demo shows the same concepts against a live sample workspace.
The definition above is the one the product uses. It is read directly from the same source the dashboard and the MCP server read, so this page cannot drift from the software.
See your evidence