- name
SKILL.md
스킬 파일은 게시자의 저장소에 그대로 있습니다.
복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add OyadotAI/oya-browser --skill codebase-design -a cursor -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add OyadotAI/oya-browser --skill codebase-design -a claude-code -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add OyadotAI/oya-browser --skill codebase-design -a codex -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add OyadotAI/oya-browser --skill codebase-design -a grok -y복사한 메시지를 Grok Bot 채팅에 보냅니다. Bot이 GitHub에서 이 SKILL.md를 가져와 모든 Bot이 공유하는 스킬에 추가합니다.
Install this skill: https://github.com/OyadotAI/oya-browser/tree/HEAD/.agents/skills/codebase-design게시자 제공
Design substantial behavior behind a small, understandable interface. Follow the project's instructions, glossary, and existing primitives. Preserve the assignment's scope and authority: a design assignment produces a proposal; implementation requires an implementation assignment.
Read the relevant callers, implementation, tests, and architectural decisions. Identify what each caller must know: types, invariants, ordering, errors, required configuration, resource ownership, and performance guarantees. Treat all of these as part of the interface. A simple signature that hides undocumented ordering requirements still has a complex interface.
Describe a concrete caller journey and where knowledge is repeated. Finish this step with source locations and an inventory of the facts callers currently carry.
Evaluate capability per fact a caller must learn, rather than implementation line counts. Concentrate decisions and failure handling where a fix can apply to all callers. Use the deletion thought experiment: if removing an abstraction spreads complexity across callers, it earns its place; if the complexity disappears, investigate whether it is only forwarding. Retain forwarding required for compatibility, policy, or ownership when those purposes are real.
Distinguish the public interface from internal composition. Keep internal helpers private when callers have no reason to know them. Use the project's terms; introduce new vocabulary only when it makes a concrete distinction clearer.
A seam is a place to change behavior without changing its caller; an adapter supplies behavior there. Justify a seam with actual variation, such as production I/O and a realistic test adapter. Keep pure calculations separate from effects where useful. Accept dependencies at an ownership boundary instead of constructing hidden global resources, without adding interfaces to every function mechanically.
For consolidation or dependencies across storage and network layers, read references/deepening.md. For a consequential design choice with credible alternatives, read references/alternatives.md.
Provide the proposed interface, one representative caller, hidden behavior, dependency owners, failure paths, compatibility impact, and migration order. Explain which existing primitives are reused. Account for cancellation, cleanup, concurrency, and rollback where relevant.
Describe verification through the interface that callers actually use. Preserve distinct behavioral coverage during a refactor; identify any adapter or integration behavior a fake cannot prove. Run or write tests only within the user's authorization. Distinguish code-based reasoning from executed evidence, and state what remains uncertain.
Complete when the recommendation explains how callers become simpler, where maintenance concentrates, and how the proposed contract can be checked. Use the team technical-plan asset when assigned as principal. Keep the proposal proportional to the change.
Adapted from Matt Pocock's engineering skills; see SOURCE.md and LICENSE.
This skill should be used when designing a module interface, reducing coupling, choosing dependency seams, improving testability, or comparing architecture alternatives.
---name: codebase-designdescription: This skill should be used when designing a module interface, reducing coupling, choosing dependency seams…---# .agents/skills/codebase-design/SKILL.mdFAQ
게시자에게서 .agents/skills/codebase-design/SKILL.md를 다운로드하거나 경로를 복사해 저장소를 엽니다. PlugMCP는 스킬을 목록에 올릴 뿐 파일을 호스팅하지 않습니다.
게시자가 https://github.com/OyadotAI/oya-browser에서 관리합니다.
Claude, Cursor, Codex, Grok Build는 게시자의 저장소에서 추가한 SKILL.md를 읽을 수 있습니다. Grok Bot은 채팅에서 스킬의 GitHub 폴더 링크를 보내면 설치합니다.