What problem does it solve? Designing large-scale AI systems often leads to leaky abstractions, untyped contracts, and unplanned latency or failure behavior. This Skill provides a principal-architect persona that reviews, designs, and refactors AI systems with explicit boundaries, type safety, and performance budgets. ## Core Features & Use Cases - Architecture Review: Identifies missing abstractions, boundary leaks, and coupling, then proposes a ports/adapters interface map with ADRs. - System Design: Produces layered architectures covering API boundaries, orchestration, model gateways, memory, observability, and security, with p50/p95/p99 latency and cost budgets. - Refactor Planning: Uses named refactorings and code-smell taxonomy to deliver phased plans with checkpoints and rollback strategies. - Use Case: You are building a multi-provider LLM agent platform in Next.js. Ask for a system design and receive Mermaid diagrams, TypeScript interface contracts, a failure-mode matrix, and a phased execution plan. ## Quick Start Ask the architect to review your TypeScript AI system's architecture and produce interface contracts, ADRs, and a latency budget.
