A `tool_use` JSON schema requires an `age` field as an integer. Can Claude return semantically implausible values like `250` or `-3` for this field?

Question

Accepted Answer

B. Yes — `tool_use` schemas enforce structural correctness (data type, field presence) but not semantic validity (plausible values). `tool_use` schemas are syntactic validators: they guarantee that the `age` field will be an integer (not a string or null). They do not enforce semantic constraints like "age must be between 0 and 130." Semantic validation — checking that integer values are within plausible human ranges, that entity types match business rules, etc. — must be implemented separately in application code. The three-layer approach is: schema enforcement (syntactic) + programmatic validation (semantic) + retry loop.

A `tool_use` JSON schema requires an `age` field as an integer. Can Claude return semantically implausible values like `250` or `-3` for this field?

Related Questions

Discussion

A tool_use JSON schema requires an age field as an integer. Can Claude return semantically implausible values like 250 or -3 for this field?

Related Questions

Discussion

A `tool_use` JSON schema requires an `age` field as an integer. Can Claude return semantically implausible values like `250` or `-3` for this field?