SKILL

check-architecture

Survey existing code and infra for deepening opportunities, places where a shallow module (an interface nearly as complex as what it hides) should become a deep one, so changes, bugs, and tests concentrate in one place. It scans the hot spots from git history, the area you name, or what a plan will touch (`for <plan>`), including pass-through Terraform modules, copied stacks, and scripts with hidden ordering. Sonnet explorers find candidates, the main agent proves each one (callers, interface size, the deletion test, and narrow `explain-why` on old decisions), and `deep` adds an Opus judge. The terminal report has at most 5 candidates with strength badges and ASCII before/after diagrams, and it can serve an HTML page on the LAN. Then one round of questions on the candidate you pick, and a hand-off to `design`. Never edits code. Triggers on "/check-architecture", "where should we refactor", "find shallow modules", "how can we make this change easy", or "architecture review of this repo"

原文 · 译文尚未补齐

适用对象:设计与内容创作。自动收录 · 待核查。

在 Codex 中使用

该组合的专门教程待补充,尚未核实适配。

准备条件

  • 本条仅收录原作者的技能元数据;请阅读完整说明,确认依赖、许可与使用条件。

选择已有说明的 AI 工具,或在页面加载后保存待办及提出教程请求。

查看原作者说明 ↗