model

Load the triad-schema skill for authoritative rules on t.model vs t.value, derivation operators, and .storage() hints.

Add a new model or value object to the project based on the user's description ($ARGUMENTS).

Steps

1. Decide model vs value object. Entities with identity → t.model. Comparable-by-value shapes (Email, Money, DateRange) → t.value. If unsure, default to t.model.

2. Place the file in src/schemas/<resource>.ts. If a schemas directory doesn't exist, create it.

3. Include derivations when appropriate:

   • CreateX = X.pick(...).named('CreateX') — omit identity/audit fields, keep user-supplied fields
   • UpdateX = X.pick(...).partial().named('UpdateX') — make every updatable field optional

4. Identity field should have both .identity() and .storage({ primaryKey: true }) if the model will become a DB table:

   id: t.string().format('uuid').identity().storage({ primaryKey: true })
   

5. Add .storage() hints to persistent fields per the triad-drizzle skill: indexed, unique, references, defaultNow, defaultRandom, columnName.

6. Update the bounded context's models[] in src/app.ts so triad validate recognizes the model as part of the context's ubiquitous language.

7. Run triad validate to confirm the new model doesn't break any consistency checks.

8. If the project uses persistence (src/db/schema.generated.ts exists), run triad db generate to regenerate the Drizzle schema.

Rules

• Always .named('NewName') when deriving. Unnamed derivations collide with the parent in OpenAPI.
• Never define a model inside a test file — tests import the real model.
• Value objects (t.value) cannot be used with body matches assertions — only models can.
• Use t.datetime() rather than t.string().format('date-time').
• Use t.empty() only for response slots (204/205/304), never for fields.

After adding the model, print the diff summary and suggest a follow-up: "Add an endpoint that uses <Model> via /triadjs:endpoint."