What problem does it solve? Teams often jump straight to building an agent with a chosen framework without a written justification for why that architecture pattern fits the problem, leading to over-engineered agent loops where a fixed workflow or plain RAG would suffice. ## Core Features & Use Cases - Structured architecture interview: Runs a batched, numbered questionnaire covering agent justification, knowledge needs, autonomy shape, workload shape, and scale constraints, each with concrete options and recommendations. - Deterministic decision logic: Maps interview answers to one of seven patterns (deterministic code, fixed workflow, RAG assistant, Agentic RAG, bounded ReAct agent, planner-executor, supervisor/multi-agent), with a mandatory human-governance wrapper for high-impact or irreversible actions. - Written design document: Produces architecture_design.md recording the chosen pattern, rejected alternatives with reasons, tool sourcing via MCP-first lookup, non-functional budgets, and evaluation gates. - Use Case: A user says "build me a LangGraph agent for customer support." The skill interviews them, determines the task is actually a fixed retrieve-then-answer flow, and documents a RAG assistant design with the agent alternative explicitly rejected. ## Quick Start Ask the agent to help design the architecture for a new agent idea and answer its numbered interview questions about your business problem and constraints.
