What problem does it solve? Turning a short product or system idea into a credible presales architecture usually requires hours of requirements gathering and diagramming, and unknowns often get presented as facts. This Skill runs a structured discovery interview, separates confirmed requirements from assumptions and open questions, and produces a decision-ready High-Level Design (HLD) document in Ukrainian with a Mermaid diagram. ## Core Features & Use Cases - Structured discovery interview: Asks one question at a time across six categories (business outcome, users, functional scope, scale, data/integrations, security) and labels every claim as Confirmed, Assumption, or Unknown before writing requirements.md. - Vendor-neutral HLD generation: Produces hld.md with one baseline architecture, at most two conditional alternatives, a claim ledger, risks, and a requirements-to-design traceability table, using a catalog of logical components (API Gateway, Event Bus, Vector DB, Feature Store, etc.). - Mermaid diagram composition rules: Enforces a readable 8-16 box diagram with 3-5 subgraphs, left-to-right happy path, and limited exception branches, plus an architecture review checklist before finalizing. - Use Case: A presales engineer receives a one-paragraph idea for a RAG knowledge assistant. The Skill interviews them about scale, access control, and integrations, then outputs a Ukrainian HLD with a rights-filtered retrieval architecture diagram and open questions affecting the estimate. ## Quick Start Ask the assistant to use the ml-system-hld skill to interview you about your system idea and produce a vendor-neutral high-level design document in Ukrainian.
