What problem does it solve? Backend features often get coded before the architecture is decided, leading to rework, unreviewed trade-offs, and design decisions lost in chat history. This Skill produces a staff-grade, 12-section system-design RFC for a new service, endpoint, or backend feature and persists it as BACKEND_SYSTEM_DESIGN.md so the architecture is reviewable before any slicing or implementation. ## Core Features & Use Cases - 12-Section Design Document: Covers problem statement, requirements, architecture, data model, API contract, caching, scaling, reliability and SLOs, security, observability, rollout, and trade-offs. - Capability-Routed, Stack-Agnostic: Designs for the stack the task actually uses (monolith, service, queue worker, serverless) read from task memory, never assuming a database or framework. - Decision and Evidence Grounding: Respects locked decisions in DECISIONS.md, marks ungrounded external contracts, and refuses to assert scale or SLO numbers without a cited source. - Use Case: Before building a new API endpoint, run this Skill to produce a complete design RFC covering the data model, failure modes, and rollout plan, then hand off to implementation-plan for slicing. ## Quick Start Ask the AI to run backend-system-design for the active task to produce the 12-section architecture RFC before any code is written.
