{
  "specVersion": "1.0",
  "host": {
    "displayName": "reopt architecture",
    "identifier": "did:web:architecture.reopt.ai",
    "url": "https://architecture.reopt.ai/",
    "description": "An AI-native system design methodology published by reopt Inc.: a four-layer model, the OCLS operating loop, eight architecture patterns, a GOVERNANCE.md specification, and installable Claude Code skills. Free to read, no account required."
  },
  "entries": [
    {
      "identifier": "urn:air:architecture.reopt.ai:index:llms-txt",
      "displayName": "reopt architecture — agent index",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/llms.txt",
      "description": "llms.txt index of the whole site, including a when-to-use section that names the jobs this methodology is right for. Start here to find any other resource.",
      "tags": [
        "index",
        "llms-txt",
        "documentation"
      ],
      "representativeQueries": [
        "what does reopt architecture cover",
        "when should I use reopt architecture"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:doc:pattern-catalog",
      "displayName": "reopt architecture — pattern catalog",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/patterns.md",
      "description": "Eight architecture patterns for AI agent products, grouped by responsibility, collaboration, and governance. Each pattern states the judgment question it answers, its forces, its solution, and the anti-pattern.",
      "tags": [
        "patterns",
        "architecture",
        "ai-agents"
      ],
      "representativeQueries": [
        "which pattern fixes unclear ownership between agents",
        "architecture patterns for multi-agent products"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:doc:governance-spec",
      "displayName": "GOVERNANCE.md specification",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/governance.md",
      "description": "A one-page structural contract for an AI product: canonical section order (OWN, CONTRACT, LAYER, SHARPEN), machine-readable frontmatter, and seven lint rules a reviewer or CI can enforce.",
      "tags": [
        "governance",
        "specification",
        "lint",
        "compliance"
      ],
      "representativeQueries": [
        "how do I write a GOVERNANCE.md for an AI product",
        "lint rules for agent governance documents"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:doc:four-layer-model",
      "displayName": "Four-layer architecture model",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/architecture.md",
      "description": "Module, Agent, Collaboration, Governance — the four layers, what question each answers that the previous cannot, plus the four types of agentic debt and the design principles behind them.",
      "tags": [
        "architecture",
        "model",
        "agentic-debt"
      ],
      "representativeQueries": [
        "how should I layer an AI agent system",
        "what is agentic debt"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:doc:glossary",
      "displayName": "reopt architecture glossary",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/glossary.md",
      "description": "Definitions this methodology relies on — agent, contract, operating loop, agentic debt, OCLS — so an agent and a team use the same vocabulary.",
      "tags": [
        "glossary",
        "terminology",
        "definitions"
      ],
      "representativeQueries": [
        "what does OCLS mean",
        "define agentic debt"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:doc:design-checklist",
      "displayName": "Agent system design checklist",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/checklist.md",
      "description": "Design-review questions for an agentic AI product, organised by architecture layer and OCLS phase. Usable in a design review, at sprint start, or in a post-incident retrospective.",
      "tags": [
        "checklist",
        "design-review",
        "quality"
      ],
      "representativeQueries": [
        "design review checklist for an AI agent system",
        "what to check before shipping an agent feature"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-responsibility-partitioning",
      "displayName": "reopt-responsibility-partitioning",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-responsibility-partitioning.md",
      "description": "여러 에이전트가 협업할 때 책임 경계를 분할하는 패턴을 적용한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "own",
        "responsibility-partitioning"
      ],
      "representativeQueries": [
        "install the reopt-responsibility-partitioning skill",
        "apply the 책임 분할 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-module-contract",
      "displayName": "reopt-module-contract",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-module-contract.md",
      "description": "모듈의 입출력·거절 조건·실패 모드를 계약으로 선언한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "contract",
        "module-contract"
      ],
      "representativeQueries": [
        "install the reopt-module-contract skill",
        "apply the 모듈 계약 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-context-routing",
      "displayName": "reopt-context-routing",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-context-routing.md",
      "description": "각 주체에 필요한 컨텍스트만 전달되도록 흐름을 설계한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "layer",
        "context-routing"
      ],
      "representativeQueries": [
        "install the reopt-context-routing skill",
        "apply the 컨텍스트 라우팅 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-state-and-memory-control",
      "displayName": "reopt-state-and-memory-control",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-state-and-memory-control.md",
      "description": "단기 상태와 장기 기억을 분리해 정보 누수와 혼선을 방지한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "layer",
        "state-and-memory-control"
      ],
      "representativeQueries": [
        "install the reopt-state-and-memory-control skill",
        "apply the 상태와 기억 제어 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-evaluation-and-guardrails",
      "displayName": "reopt-evaluation-and-guardrails",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-evaluation-and-guardrails.md",
      "description": "허용 판단과 위험 판단을 평가 기준과 안전 규칙으로 구분한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "sharpen",
        "evaluation-and-guardrails"
      ],
      "representativeQueries": [
        "install the reopt-evaluation-and-guardrails skill",
        "apply the 평가와 가드레일 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-human-approval",
      "displayName": "reopt-human-approval",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-human-approval.md",
      "description": "고비용·고위험·고영향 결정을 인간 승인 흐름 안에 둔다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "sharpen",
        "human-approval"
      ],
      "representativeQueries": [
        "install the reopt-human-approval skill",
        "apply the 인간 승인 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-cost-control",
      "displayName": "reopt-cost-control",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-cost-control.md",
      "description": "토큰 예산·모델 선택·호출 빈도를 구조적으로 관리해 비용을 통제한다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "layer",
        "cost-control"
      ],
      "representativeQueries": [
        "install the reopt-cost-control skill",
        "apply the 비용 통제 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:reopt-decision-traceability",
      "displayName": "reopt-decision-traceability",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/reopt-decision-traceability.md",
      "description": "판단 근거·선택 사유·협업 경로를 구조화된 로그로 남긴다. Copy into .claude/skills/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "pattern",
        "sharpen",
        "decision-traceability"
      ],
      "representativeQueries": [
        "install the reopt-decision-traceability skill",
        "apply the 의사결정 추적 pattern in Claude Code"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:ocls-review",
      "displayName": "/ocls-review",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/ocls-review.md",
      "description": "현재 변경사항을 OCLS 4단계로 검증하고 위반 항목의 패치 제안을 생성한다. Copy into .claude/commands/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "command"
      ],
      "representativeQueries": [
        "install the /ocls-review skill"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:debt-audit",
      "displayName": "/debt-audit",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/debt-audit.md",
      "description": "최근 스프린트의 부채 4종을 감사하고 GOVERNANCE.md 갱신안을 생성한다. Copy into .claude/commands/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "command"
      ],
      "representativeQueries": [
        "install the /debt-audit skill"
      ]
    },
    {
      "identifier": "urn:air:architecture.reopt.ai:skill:governance-reviewer",
      "displayName": "governance-reviewer",
      "type": "text/markdown",
      "url": "https://architecture.reopt.ai/ko/skills/governance-reviewer.md",
      "description": "PR·브랜치 변경을 OCLS·8패턴 관점으로 리뷰하는 sub-agent. Copy into .claude/agents/ of a Claude Code project to activate it.",
      "tags": [
        "claude-code",
        "skill",
        "agent"
      ],
      "representativeQueries": [
        "install the governance-reviewer skill"
      ]
    }
  ]
}