7. Validate input with constraints

You want the schema to enforce shape — a pattern on an ID, a length range on a reason, a numeric bound — so the LLM gets it right the first time and bad inpu...

Problem. You want the schema to enforce shape — a pattern on an ID, a length range on a reason, a numeric bound — so the LLM gets it right the first time and bad input never reaches call.

Solution. Stack constraint tags after the type. They're type-aware: @min/@max become minLength, minimum, or minItems depending on the field's type.

#: (
#:   applicant_id: String   @pattern(^app-\d+$) @desc(Use fetch_latest_applicant to get this ID),
#:   workflow_id: String    @min(1) @max(50),
#:   email: String          @format(email),
#:   priority: Integer      @min(1) @max(5),
#:   ?reason: String?       @min(10) @max(500),
#:   ?tags: Array[String]   @max(10) @unique()
#: ) -> Hash[Symbol, untyped]

Result. Each tag maps to its JSON Schema keyword: pattern, minLength/maxLength, format, minimum/maximum, maxItems, uniqueItems. @desc(...) becomes the field description — and doubles as a tool-chaining hint to the LLM ("call X first to get this value").

Full tag table lives in the README. The common ones: @min @max @exclusive_min @exclusive_max @multiple_of @pattern @format @unique @desc @title @example @deprecated.

Collected from COOKBOOK.md in the repository. Edit it there, not here.