What problem does it solve? AI code generators produce syntactically correct code that is often architecturally unsound, with unclear state ownership, missing observability, and hidden failure modes. This Skill enforces design-first thinking so systems are specified, auditable, and resilient before and after code generation. ## Core Features & Use Cases - Three Pillars Framework: Audits any system for state ownership, feedback/observability, and blast radius using structured checklists. - Spec and Design Templates: Provides an architectural spec template and a Google DESIGN.md template for visual design tokens with WCAG validation. - Code Review Checklist: Audits AI-generated code across spec compliance, error handling, concurrency, security, and performance. - Automation Scripts: Includes branerail-audit to verify project structure (CLAUDE.md, /specs, DESIGN.md) and branerail-chaos to generate chaos-test prompts from spec failure modes. - Use Case: Before building a checkout service, write a spec from the template, prompt Claude Code against it, then run the review checklist to catch missing circuit breakers and untested failure modes. ## Quick Start Ask the agent to review your system design or AI-generated code against the Three Pillars of state, feedback, and blast radius.