What problem does it solve? Answers about OpenAI products, Codex behavior, model pricing, and API usage are often based on stale memory or unreliable snippets. This Skill enforces a docs-first workflow that searches, fetches, and cites official OpenAI documentation before answering, so responses reflect current product reality. ## Core Features & Use Cases - Official-source enforcement: Searches and fetches only developers.openai.com, platform.openai.com, and learn.chatgpt.com, citing the exact page behind each claim. - Routed workflows: Dedicated routes for model selection, model migration, prompting guidance, ChatGPT Work comparisons, local documentation MCP setup, and broad Codex self-knowledge via a cached manual. - Helper scripts: Includes a Codex manual fetcher with hash-verified caching and a latest-model resolver that returns the current model plus migration and prompting guide URLs. - Use Case: When asked to migrate an integration to the latest OpenAI model, the Skill resolves the current target model, fetches the live migration and prompting guides, and preserves any explicitly requested model instead of substituting a newer one. ## Quick Start Ask the agent to look up the current official OpenAI documentation for a Codex setting, model, or API feature and answer with citations.