bd-attribute

better-data: Adding a new attribute

For library maintainers introducing a new declarative hint that DTO authors will place on constructor parameters / properties — #[ArrayOf], #[Default], #[ListOf]-style markers, domain-specific decorators. The attribute itself is just a data carrier; the work is wiring it into every engine that reads attributes, because partial wiring silently degrades.

Misconception this skill corrects

"I'll add src/Attribute/Foo.php and read it in PostSink — done."

Attributes are read by multiple engines, and missing one leaves a feature that works on a happy path and fails subtly elsewhere. The original encrypt flag was meta-only; when options needed the same semantics, the partial wiring left a footgun until #[Encrypted] replaced it across OptionSink, PostSink, AttributeDrivenHydrator, RestSchemaBuilder, and Presenter — all in one go.

The engines that read attributes today, all need to know about a new attribute relevant to their concern:

• Read-side hydration: AttributeDrivenHydrator (src/Internal/AttributeDrivenHydrator.php) and DataObject::coerceParameter (src/DataObject.php:168).
• Write-side projection: SinkProjection::prepareValue (src/Internal/SinkProjection.php:193) and OptionSink::projectForStorage (src/Sink/OptionSink.php:119).
• REST / OpenAPI schema: RestSchemaBuilder::buildProperty (src/Internal/RestSchemaBuilder.php).
• Output / display: Presenter::sensitiveFieldNames and friends in src/Presenter/Presenter.php.

If your attribute is a write-time concern (encryption, formatting, slashing), all four still need to coordinate — read-side has to invert what write-side did.

When to use this skill

Trigger when ANY of the following is true:

• Creating a new file under src/Attribute/.
• Adding #[Attribute(...)] to any class.
• The diff adds a new attribute reference (#[NewThing]) to a DTO and to the consumer engine.
• Reviewing a PR that wires an attribute into ONE engine — flag every other relevant engine as missing.