What is ai-rag-agent-designer-skill?

AI RAG and Agent Designer Skill is a Codex skill aimed at AI product managers, meant to help decide whether an AI product actually needs RAG (retrieval augmented generation, a way of pulling in outside knowledge for an AI to use) or an autonomous agent that can use tools, and to produce a design that can be pasted straight into a PRD, or product requirements document. The README describes several situations where it is meant to help: when you have an AI product idea but are not sure if RAG is needed, when you have a raw PRD and need to work out how the knowledge base, retrieval, agent, tools, and memory should be designed, when you are preparing for an AI product manager interview and need to explain RAG and agent design clearly, or when you need to turn a business requirement into something engineers and AI practitioners can understand. You trigger it with plain requests such as asking it to design the RAG and agent plan for a product, or asking whether a given scenario needs RAG or an agent at all. Its workflow first classifies your input, then asks clarifying multiple choice questions if the information is unclear, then restates the underlying business problem. From there it separately judges whether RAG is needed, and if so designs the data cleaning, chunking, indexing, retrieval, reranking, source tracing, evaluation, and fallback behavior, and whether an agent is needed, and if so designs the agent type, responsibilities, tools, memory, prompts, risk handling, evaluation, and fallback behavior. It finishes by producing a Markdown plan meant to be usable directly inside a PRD. The stated design principles are that it does not assume every AI product needs RAG or an agent, that it asks the user to choose rather than guessing when information is missing, and that its output favors plain product language over technical jargon. The repository is small, holding just the skill definition file, an agent metadata file, and a references folder with separate write-ups for RAG decisions, agent decisions, and the output template. Installing it means cloning the repository into a local Codex skills folder. The README does not state a license for the code.