Glossary

AI Readiness

A measurement of whether an AI agent can discover, understand and correctly use a design system, scored across eight weighted dimensions.

Readiness is a property of the system, not of the model. It asks whether the information an agent needs is written down anywhere it can read: what a component is for, when not to use it, which props are valid, which Figma component became which implementation. The score summarises that; the findings behind it are what you act on.

System Graph

Onefold's structured representation of a design system, as typed nodes such as components, tokens and stories, and typed relationships between them.

The graph is what makes cross-source questions answerable. A component node can carry edges to the token it consumes, the story that documents it, the Figma component it implements and the documentation that describes it, regardless of which source each fact came from.

Source

Anything connected to Onefold that facts are read from, such as a GitHub repository, a Figma library or a Storybook.

A source is the unit of connection and of deletion. Disconnecting one removes everything derived from it, which is what keeps provenance honest: no assertion outlives the source it came from.

Observation

Something discovered from a source, kept with its provenance, before it becomes canonical knowledge in the System Graph.

Observations are raw and may conflict. Two sources can disagree about the same component, and both observations survive with their evidence, rather than one silently overwriting the other on the way into the graph.

Mapping

A relationship between a design component and its code implementation, carrying the evidence behind it and a confidence value.

Mappings are the hardest part of the product and the most useful. A name match is weak evidence, a name match plus matching variants plus a documented reference is strong evidence, and the difference is recorded rather than rounded to a yes.

Agent Context

The retrieval-oriented representation of the System Graph that AI agents consume, whether as generated files or over MCP.

Agent context is not documentation with the prose removed. It is the subset of the graph that answers the questions an agent actually asks while writing code: what exists, what is this for, when should it not be used, what does correct usage look like.

System Rule

An explicit constraint a design system places on its own use, such as a deprecation, a required pairing or a forbidden combination.

Rules are the part of a design system that lives in people's heads and in review comments. Written down, they are the difference between an agent picking a plausible component and picking the right one.

Design System Drift

Divergence between the documented system, the design files and the implemented code.

Drift is normal and continuous. It matters more with agents in the loop, because a human notices that a Figma variant does not exist in code while an agent simply generates the variant that does not exist.

The model behind the first term is described in full on AI Readiness, and the practical version is in the help center.