What problem does it solve? Choosing the right architecture for an AI agent system is hard: developers must pick among many agentic design patterns, combine them correctly, weigh latency/cost/autonomy trade-offs, and translate the design into concrete platform primitives. This Skill guides that design phase so agent systems are grounded in proven patterns rather than ad-hoc prompts. ## Core Features & Use Cases - Pattern selection and composition: Recommends from 21 agentic design patterns (routing, RAG, reflection, guardrails, HITL, and more) with selection tables, troubleshooting guides, and six ready-made composition recipes. - Platform mapping: Maps every pattern to concrete Resparkable orchestration primitives such as agents, capabilities, workflow step types, knowledge bases, and evaluation judges. - Production readiness review: Provides trade-off analysis, anti-pattern warnings, and a production checklist covering rate limits, idempotency, observability, and cost alerting. - Use Case: A developer says "build me a support chatbot that looks up orders and escalates refunds." The Skill recommends Routing + RAG + Tool Use + Guardrails + HITL, maps them to a route step, knowledge base, capabilities, and a human_approval step, then hands off to the orchestration-solution-builder for implementation. ## Quick Start Ask the agent architect to design an AI agent system for your use case, for example by describing the business problem and requesting a pattern-based architecture mapped to Resparkable primitives.
