What problem does it solve? Architecture boundaries erode over time as codebases grow: frontend code starts importing backend internals, secrets leak into client bundles, and route handlers bypass service layers. This Skill turns existing architecture documentation and directory structure into concrete, executable boundary checks before drift becomes a production incident. ## Core Features & Use Cases - Evidence-Based Check Proposals: Generates boundary checks only from observed evidence in docs/ARCHITECTURE.md, root instructions, directory layout, and existing lint/test/CI configuration. - Common Boundary Patterns: Covers cross-layer import prevention, server-only secret isolation, route-to-service separation, domain independence from UI, and shared package purity. - Approval-Gated Script Generation: Outputs a structured proposal (boundary, evidence, script target, false-positive risk) and never writes check scripts without explicit approval. - Use Case: A team notices a client bundle accidentally imported a server-only module. Use this Skill to inspect the repository layout and propose a CI check that blocks browser code from importing server internals. ## Quick Start Ask the agent to analyze this repository's architecture docs and directory layout and propose executable boundary checks for cross-layer imports and secret leaks.
