<skill_overview> <purpose>Serialize and deserialize Rust types safely and predictably</purpose> <triggers> <trigger>Designing request/response DTOs</trigger> <trigger>Parsing JSON or config files</trigger> <trigger>Customizing field names or defaults</trigger> <trigger>Handling versioned schemas</trigger> </triggers> <sources> <source url="https://github.com/serde-rs/serde">Serde GitHub</source> <source url="https://docs.rs/serde">Serde Docs</source> </sources> </skill_overview> <derive> <rules> <rule>Use #[derive(Serialize, Deserialize)] for DTOs</rule> <rule>Keep DTOs separate from domain entities</rule> <rule>Prefer strongly typed structs over Value</rule> </rules> </derive> <field_attributes> <rules> <rule>Use #[serde(rename_all = "camelCase")] for JSON APIs</rule> <rule>Use #[serde(default)] for optional fields</rule> <rule>Use #[serde(skip_serializing_if = "Option::is_none")] for omitting nulls</rule> <rule>Use #[serde(flatten)] to merge nested fields</rule> </rules> <example> <code> #[derive(Serialize, Deserialize)] #[serde(rename_all = "camelCase")] struct UserDto { #[serde(default)] display_name: String, #[serde(skip_serializing_if = "Option::is_none")] email: Option<String>, } </code> </example> </field_attributes> <validation_and_schema> <rules> <rule>Use #[serde(deny_unknown_fields)] for strict inputs</rule> <rule>Version payloads via enums or tagged variants</rule> </rules> </validation_and_schema> <custom_serialization> <rules> <rule>Use serialize_with/deserialize_with only when necessary</rule> <rule>Prefer newtype wrappers for custom formats</rule> </rules> </custom_serialization> <anti_patterns> <avoid name="untagged_everywhere">Avoid untagged enums for public APIs</avoid> <avoid name="stringly_typed">Avoid mapping enums to raw strings manually</avoid> <avoid name="value_for_core">Avoid serde_json::Value for core flows</avoid> </anti_patterns>