ODD-min: An Auditable Output Governance Interface for AI-Assisted Software Engineering

AI-assisted software engineering can produce code, configuration, tests, prompts, and other engineering artifacts faster than conventional review processes can inspect them line by line. This preprint introduces ODD-min, a minimal auditable output-governance interface for such artifacts. ODD-min defines five fields: contract, validation, evidence, seal, and rollback/recall. These fields produce auditable governance records with gate states such as PASS, FAIL, FREEZE, and RECALL_REQUIRED. ODD-min is not presented as a substitute for software engineering, expert review, safety certification, legal approval, or organizational governance. It is positioned as an artifact-level evidence gate that can complement requirements engineering, CI/CD evidence, safety or assurance cases, AI risk management frameworks, and organizational AI management systems. The accompanying artifact provides a draft JSON schema, representative records for four gate outcomes, and a minimal dependency-free reference implementation. The preprint makes a deliberately limited claim: ODD-min can represent auditable governance records for AI-assisted engineering artifacts, the representative records can be checked for required fields and expected gate states, and a stable hash-only seal can be generated for the same representative record. It does not establish that ODD-min improves software quality in industrial settings, substitutes for human review, or guarantees safety, compliance, or business correctness.