What problem does it solve? When changing runtime behavior, exported APIs, serialized state, or configuration surfaces in the openai-agents-python SDK, it is easy to either break released contracts or over-engineer shims for unreleased code. This Skill provides a decision framework for judging breaking-change risk against the latest release tag before editing code. ## Core Features & Use Cases - Compatibility boundary analysis: Classify changes as released public API, unreleased branch-local interfaces, internal helpers, or durable external state, and decide whether shims, migrations, or direct rewrites are warranted. - Release-tag grounding: Resolve the latest release tag from the origin remote (falling back to local tags) so breaking-change judgments are based on shipped contracts, not branch churn. - SDK-specific rules: Guidance for OpenAI API feature gaps, streaming vs non-streaming parity, public dataclass field ordering, strict-mode opt-ins, and None semantics for configuration. - Use Case: Before refactoring a session persistence API on a feature branch, use this Skill to confirm the interface is unreleased, then rewrite it directly without adding aliases or deprecation shims. ## Quick Start Use the implementation-strategy skill to decide the compatibility boundary and implementation approach before I edit this runtime code.
