domain-model

Read the intent spec, identify primary subjects as candidate entities and recurring patterns as candidate concepts, then classify each against the existing model. Write or update files under docs/domain-model/concepts/ and docs/domain-model/entities/, update both index files, and link the model back into the spec.

An entity is a thing with its own identity persisted in its own database table, like an item or wear log. A concept is an abstract recurring pattern or mechanism, like status markers or filtering, that appears across multiple features and is instantiated as fields on entities.

No. Junction tables exist only to relate two entities, so they are documented under the Relationships section of the entities they link and listed in the relationships table of the entities index. Only strong, independent tables qualify as entities.

Re-extract whenever the intent spec changes, before any code implementing the new behavior is written. Update all affected concept and entity files, both index files, and the spec's Domain Model section so the plan derives from the current model.

Skip extraction for changes with no domain impact, such as pure code refactors that leave the spec, concepts, and entities unchanged. In that case, note in the implementation plan that the domain model is unchanged.