schema-hardening

name	schema-hardening
description	Expert review of Pydantic and data transfer schema validation constraints to prevent mass assignment vulnerabilities.
skill_type	universal
version	1.0.0

Schema Hardening & Input Validation

Expert guidelines to check input schemas (Pydantic models, JSON schemas) and prevent mass-assignment (CWE-915) or field-injection vulnerabilities.

🛡️ Hardening Rules

• Mass Assignment Prevention (CWE-915): All public request-facing schemas must prevent extra/unexpected payload attributes. In Pydantic v2, declare model_config = ConfigDict(extra="forbid") or inherit from a project-wide hardened base class.
• Strict Validation: Always validate datatypes strictly (e.g. use specific types like EmailStr, PositiveInt, Field(max_length=...) instead of raw str or int attributes).
• Safe Fallbacks: Enum fields must validate against actual enum classes, preventing bypass of custom fallback logic in schemas.

🚫 Rationalisations to Reject (Anti-Rationalisation)

Excuse / Rationalisation	Why it fails / Rebuttal
"Pydantic ignores extra parameters by default, so we don't need extra='forbid'."	Ignoring parameters allows clients to send unvalidated inputs that could silently bind to backend models or logic. Hard-failing is the safe poka-yoke design.
"I will add validation logic directly inside the repository layer later."	Input schemas at the boundary must validate payloads first before they reach any service or database layer.