{"_id":"@dusky-bluehour/agent-service","_rev":"7-b7d9b60a6258480b9f0869ee030b3ad9","name":"@dusky-bluehour/agent-service","dist-tags":{"latest":"0.6.8"},"versions":{"0.6.2":{"name":"@dusky-bluehour/agent-service","version":"0.6.2","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.2","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-os":"scripts/init.mjs","tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"ca0d41e314c4162a574073ef1a6f7fe6adce2dd2","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.2.tgz","fileCount":64,"integrity":"sha512-wBWIj63giq6j52/unX6XCAYpdOMeqWePlpxxBJQlYK8ii6Rigns+u9qkSRjO/j4Zt5JVOtSzKGBolpOEjKEK5g==","signatures":[{"sig":"MEUCIBdbSGVp4hMqLE93i9855Ch1GvkrPJnzyWmfl60668DJAiEAloWSNfP4X7BHoVjibfZHoCYNiOKUM2qCY6wH9oLm8Cw=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":227681},"type":"module","engines":{"node":">=18"},"scripts":{"validate":"node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.2_1771411096496_0.2791576299078664","host":"s3://npm-registry-packages-npm-production"}},"0.6.3":{"name":"@dusky-bluehour/agent-service","version":"0.6.3","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.3","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-os":"scripts/init.mjs","tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"486b4fff5839e429d1d9eb5c5a9ded76845da062","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.3.tgz","fileCount":64,"integrity":"sha512-OG/nojNW0Q1S9KA5ivR+aGMSFVZ3EuzPa+xQ262fSKwVxPi0n7Lfp6golrTtXF3KubwZAR+EgszAjoMnqBOb6Q==","signatures":[{"sig":"MEQCIC0Hj6Tz+i7lhJ0ovzFTq4DikTRW3TeBzIDr8a78oTXCAiASBLBkSXojVbqOLUDEr8P+eV65KDvXHAqNFlvn2WSLWw==","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":249784},"type":"module","engines":{"node":">=18"},"scripts":{"validate":"node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.3_1771431513384_0.7862648740337157","host":"s3://npm-registry-packages-npm-production"}},"0.6.4":{"name":"@dusky-bluehour/agent-service","version":"0.6.4","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.4","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-os":"scripts/init.mjs","tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"2eff2ec9a24646cbf7fdbb3c90351c9643952a69","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.4.tgz","fileCount":64,"integrity":"sha512-8rNrtTZyDoFg5RYaXrRlFI5VEgmoKsad/UKLBZDUPTVGfJuLprwOn5sdgueJTap6RDoNxAv3GCNCPFlyBZMEJg==","signatures":[{"sig":"MEUCIBrpcMFNv1hIu6BnUOyONPTyx85pjU+1Z3cU+O8QJf+jAiEAoaRarXjfs65SpYrc+tC48+3xm1aXLkB72sEnPq2Cmlk=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":253568},"type":"module","engines":{"node":">=18"},"scripts":{"validate":"node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.4_1771432420775_0.3676496314305655","host":"s3://npm-registry-packages-npm-production"}},"0.6.5":{"name":"@dusky-bluehour/agent-service","version":"0.6.5","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.5","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"8c250b282205f0ea6c878446266a3254921df082","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.5.tgz","fileCount":64,"integrity":"sha512-0IjhGFugWL1Ii45cbubpv9/TRe8Wa627R672TZkuE+DWssGIexy7ebTuoLcrjJDFejekqDMJMbjIWIVN3X0E3g==","signatures":[{"sig":"MEUCIQCU7rKtr9k6MpwvLT1yvscLBycGZmsXQaSNU1AMMiUP6gIgF6m6SRfwNytfhFj3Q9WqXYhcAN//xlvDBnhLjiPPnnE=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":246828},"type":"module","engines":{"node":">=18"},"scripts":{"validate":"node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.5_1771470066163_0.06704450076117263","host":"s3://npm-registry-packages-npm-production"}},"0.6.6":{"name":"@dusky-bluehour/agent-service","version":"0.6.6","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.6","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"e25d5151a81e5d4200e3d8a20d6a5cc91e38465f","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.6.tgz","fileCount":81,"integrity":"sha512-ykV/YHjnifiUDu7bepE9OcwzOj1FF0Q36ZZMj3To51D46c+10odo5yIv5y3SwGiR0Lr1iQXgb1pjotWcKbLKYw==","signatures":[{"sig":"MEUCIQC6C3aVd6FUkUAppDTODSchVWJ2jdjSmIMyYC8bIMe8PwIgIRdZtKctg61K+T8BLoaPNs8slyibMtnTBY76zFf0FQo=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":431623},"type":"module","engines":{"node":">=18"},"gitHead":"c0e3fefd82179825dc93bd170945351ba89b4d0e","scripts":{"generate":"node scripts/generate-from-common.mjs","validate":"npm run generate:check && node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","generate:check":"node scripts/generate-from-common.mjs --check","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.6_1771491081766_0.16293499554012092","host":"s3://npm-registry-packages-npm-production"}},"0.6.7":{"name":"@dusky-bluehour/agent-service","version":"0.6.7","keywords":["ai-agent","workflow","codex","claude-code","antigravity","devops","security","code-review"],"author":"","license":"MIT","_id":"@dusky-bluehour/agent-service@0.6.7","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"bin":{"tri-agent-manager":"scripts/init.mjs"},"dist":{"shasum":"bea91eaf97507633d1248f7f598fb94526f90d3f","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.7.tgz","fileCount":169,"integrity":"sha512-enbOn/zO8dM+G+Z4G7x7KdxILuyUUXa4LLISNIKWehklSwkrTjyTAvlrmfzlnL2F0YQ3N/C+DhQhkywVFBqESw==","signatures":[{"sig":"MEUCIQDOvOpqwogThIvgv8pon9OGGjOo2L5miXKQ+pfZh+FB9AIgViB4ctu4LLDnNs3Y+vb/FBht/MF20n3K/6RP1vAIIIY=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":570104},"type":"module","engines":{"node":">=18"},"gitHead":"b6a97277b8f1ca6e19722d9441e359d98fa86bd1","scripts":{"generate":"node scripts/generate-from-common.mjs","validate":"npm run generate:check && node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","generate:check":"node scripts/generate-from-common.mjs --check","prepublishOnly":"npm run prepublish:check","prepublish:check":"npm run validate && npm run pack:dry-run"},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"_npmVersion":"11.6.2","description":"Service operation skills/workflows pack for Claude Code, Antigravity, and Codex","directories":{},"_nodeVersion":"24.13.0","publishConfig":{"access":"public"},"_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/agent-service_0.6.7_1771509569588_0.8490180460501275","host":"s3://npm-registry-packages-npm-production"}},"0.6.8":{"name":"@dusky-bluehour/agent-service","version":"0.6.8","publishConfig":{"access":"public"},"description":"Service operation skills/workflows pack for Claude Code, Gemini CLI, and Codex","type":"module","bin":{"tri-agent-manager":"scripts/init.mjs"},"scripts":{"generate":"node scripts/generate-from-common.mjs","generate:check":"node scripts/generate-from-common.mjs --check","validate":"npm run generate:check && node scripts/validate.mjs","pack:dry-run":"npm_config_cache=.npm-cache npm pack --dry-run","prepublish:check":"npm run validate && npm run pack:dry-run","prepublishOnly":"npm run prepublish:check"},"keywords":["ai-agent","workflow","codex","claude-code","gemini-cli","devops","security","code-review"],"author":"","license":"MIT","engines":{"node":">=18"},"gitHead":"2611bfe899675b51dcef2eb7928b6d40dabced5e","_id":"@dusky-bluehour/agent-service@0.6.8","_nodeVersion":"24.13.0","_npmVersion":"11.6.2","dist":{"integrity":"sha512-aCM0PL4Re7BFbNFkAB9x5JfFhbqdFugLnWsLUmHwDS25TRd28Dx/Bu/loD3Y1iyA7cwE+amiKz6qc8TQsaHeCg==","shasum":"6b01cbd0387fc628fc6fef57e94c4db27d5b6aba","tarball":"https://registry.npmjs.org/@dusky-bluehour/agent-service/-/agent-service-0.6.8.tgz","fileCount":168,"unpackedSize":556975,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEYCIQD1QDL9szV0xwOOejP5Xtmb8aiheobxLOmqZXL8mFp1WgIhALh8t+hCzURVTpS6bEyNHFpio1nDD+agsHiWwFwJG+G/"}]},"_npmUser":{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"},"directories":{},"maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/agent-service_0.6.8_1772266106739_0.2122674215904985"},"_hasShrinkwrap":false}},"time":{"created":"2026-02-18T10:38:16.425Z","modified":"2026-02-28T08:08:27.019Z","0.6.2":"2026-02-18T10:38:16.630Z","0.6.3":"2026-02-18T16:18:33.529Z","0.6.4":"2026-02-18T16:33:40.930Z","0.6.5":"2026-02-19T03:01:06.337Z","0.6.6":"2026-02-19T08:51:21.934Z","0.6.7":"2026-02-19T13:59:29.741Z","0.6.8":"2026-02-28T08:08:26.895Z"},"license":"MIT","keywords":["ai-agent","workflow","codex","claude-code","gemini-cli","devops","security","code-review"],"description":"Service operation skills/workflows pack for Claude Code, Gemini CLI, and Codex","maintainers":[{"name":"dusky-bluehour","email":"dusky.bluehour@gmail.com"}],"readme":"# tri-agent-manager\n\nClaude Code, Gemini CLI, Codex 3개 AI 에이전트용 **서비스 운영 스킬/워크플로우 설치기**입니다.\n\n공통 정의(common/) 하나로 3개 플랫폼에 동일한 운영 역량을 배포합니다.\n\n## 아키텍처 개요\n\n```\ncommon/                          ← 단일 원천 (Single Source of Truth)\n  skills/skill-catalog.json         스킬 8개 (플랫폼별 procedures 포함)\n  commands/command-catalog.json     명령 20개\n  workflows/workflow-catalog.json   워크플로우 5개 (Gemini CLI 전용)\n  settings/security-policy.json     보안 정책\n  claude/subagent-catalog.json      Claude 서브에이전트 15개\n  claude/team-catalog.json          Claude 팀 3개\n  gemini/agent-catalog.json         Gemini 에이전트 카탈로그\n  gemini/artifact-catalog.json      Gemini 아티팩트 계약\n       │\n       ▼  npm run generate\n       │\n  ┌────┴────────────────────────────────────────────┐\n  │                                                  │\nclaude-code/          gemini/              codex/\n  skills/               skills/              skills/\n  subagents/            agents/              skills/cmd-*/\n  agent-teams/          artifacts/           automations/\n  commands/native/      commands/definitions/  commands/\n    *.md (Slash Cmd)      *.toml (Extension)   command-catalog.json\n  instructions/         workflows/           instructions/\n    CLAUDE.template.md    definitions/*.yaml   AGENTS.template.md\n  settings/             gemini-extension.json  settings/\n    settings.json       instructions/\n                          GEMINI.template.md\n```\n\n## 빠른 시작\n\n```bash\n# 대화형 설치 (프리셋 선택 → 도구/구성요소 선택 → 설치)\nnpx --yes --package @dusky-bluehour/agent-service tri-agent-manager --interactive\n```\n\n```bash\n# 명시적 install / update\nnpx --yes --package @dusky-bluehour/agent-service tri-agent-manager install --interactive\nnpx --yes --package @dusky-bluehour/agent-service tri-agent-manager update --interactive\n```\n\n```bash\n# Non-interactive (CI/스크립트용)\nnpx --yes --package @dusky-bluehour/agent-service tri-agent-manager install \\\n  --preset balanced-core --target . --yes\n```\n\n## 설치하면 무슨 일이 생기나\n\n설치 대상 프로젝트에 아래 파일이 생성됩니다.\n\n```\n<project>/\n  .claude/                    ← Claude Code 선택 시\n    skills/<name>/SKILL.md       8개 스킬\n    agents/<role>.md             15개 서브에이전트\n    agent-teams/team-catalog.json\n    commands/<name>.md           20개 네이티브 슬래시 명령\n    instructions/CLAUDE.template.md\n    settings/settings.json\n  .gemini/                    ← Gemini CLI 선택 시\n    skills/<name>/SKILL.md       8개 스킬\n    agents/agent-catalog.json\n    artifacts/artifact-catalog.json\n    workflows/                   5개 워크플로우 정의\n    commands/definitions/*.toml  20개 커스텀 명령\n    gemini-extension.json        Extension 매니페스트\n    instructions/GEMINI.template.md\n    settings/editor-policy.json\n  .codex/                     ← Codex 선택 시\n    skills/<name>/SKILL.md       8개 스킬\n    skills/cmd-*/SKILL.md        20개 명령 스킬\n    automations/automation-recipes.toml\n    commands/command-catalog.json\n    instructions/AGENTS.template.md\n    settings/runtime-policy.json\n  CLAUDE.md                   ← 프로젝트 규칙 파일 (Claude Code)\n  GEMINI.md                   ← 프로젝트 규칙 파일 (Gemini CLI)\n  AGENTS.md                   ← 프로젝트 규칙 파일 (Codex)\n  .tri-agent-manager/\n    state.json                   설치 상태 기록\n    USAGE.ko.md                  실행 가이드\n```\n\n## 동작 방법\n\n### 1. 스킬 (Skills) — 3개 플랫폼 공통\n\n8개 운영 스킬이 각 에이전트의 네이티브 포맷으로 설치됩니다.\n\n| 스킬 ID | 역할 |\n|---------|------|\n| `change-safety-review` | 변경사항 회귀 검증 + 코드 리뷰 |\n| `code-review-and-improvement` | 코드 품질 개선 + 기술부채 정리 |\n| `frontend-repetition-pack` | UI 컴포넌트화 → Hook 분리 → 성능 최적화 |\n| `incident-response` | 장애 트리아지 → 근본원인 수정 → 포스트모템 |\n| `prd-to-production-pipeline` | PRD 작성 → 자동 구현 → 검증 → 배포 |\n| `release-and-operations` | CI/CD 게이트 → 배포 → 모니터링 |\n| `security-hardening` | 위협 모델링 → 취약점 감사 → 하드닝 |\n| `service-lifecycle-orchestration` | 전체 수명주기 오케스트레이션 |\n\n각 스킬은 자체적으로 실행 가능한 상세 절차를 포함합니다. 외부 명령(CMD-*) 참조 없이 절차를 직접 기술합니다.\n\n사용 방법:\n- **Claude Code**: `.claude/skills/<name>/SKILL.md`를 자동 인식하여 사용자 요청에 매칭\n- **Gemini CLI**: `.gemini/skills/<name>/SKILL.md`를 Extension 내 스킬로 로드\n- **Codex**: `.codex/skills/<name>/SKILL.md` + `agents/openai.yaml`로 스킬 UI 표시\n\n### 2. 명령 (Commands) — 플랫폼별 네이티브 변환\n\n20개 공통 명령이 각 플랫폼의 네이티브 실행 포맷으로 변환됩니다.\n\n| 플랫폼 | 변환 결과 | 실행 방식 |\n|--------|----------|----------|\n| Claude Code | `commands/native/*.md` | `/cmd-dev-be-api` 슬래시 명령으로 실행 |\n| Gemini CLI | `commands/definitions/*.toml` | Extension 커스텀 명령으로 실행 |\n| Codex | `skills/cmd-*/SKILL.md` | 스킬로 변환하여 네이티브 실행 |\n\n### 3. 워크플로우 (Workflows) — Gemini CLI 전용\n\n단계형 에이전트 체인 워크플로우는 Gemini CLI만 네이티브 지원합니다.\n\n| 워크플로우 ID | 용도 |\n|--------------|------|\n| `WF-SERVICE-E2E` | 신규 서비스 E2E (Plan→Implement→Test→Review→Release) |\n| `WF-FRONTEND-REFACTOR` | 프론트엔드 반복 개선 (UI→Hook→성능) |\n| `WF-SECURITY-HARDENING` | 보안 강화 배치 (진단→보완→검증) |\n| `WF-INCIDENT-RESPONSE` | 장애 대응 (트리아지→근본수정→포스트모템) |\n| `WF-PRD-TO-PRODUCTION` | PRD 대화형 작성 → 순차 자동 구현 |\n\n각 워크플로우는 `workflow-catalog.json`에 단계/에이전트/아티팩트 계약이 정의되어 있고, `definitions/*.workflow.yaml`로 실행합니다.\n\n### 4. 서브에이전트 & 팀 — Claude Code 전용\n\nClaude Code는 리드-전문가 구조의 서브에이전트 팀을 지원합니다.\n\n- 15개 서브에이전트: `lead-orchestrator`, `backend-engineer`, `security-engineer`, `sre-release-engineer` 등\n- 3개 팀: `TEAM-SVC-DELIVERY` (서비스 딜리버리), `TEAM-SEC-HARDENING` (보안), `TEAM-INCIDENT` (장애 대응)\n\n### 5. 프로젝트 규칙 파일\n\n설치 시 프로젝트 루트에 에이전트별 규칙 파일을 자동 생성합니다 (기존 파일 유지).\n\n| 에이전트 | 규칙 파일 | 템플릿 원본 |\n|---------|----------|------------|\n| Claude Code | `CLAUDE.md` | `instructions/CLAUDE.template.md` |\n| Gemini CLI | `GEMINI.md` | `instructions/GEMINI.template.md` |\n| Codex | `AGENTS.md` | `instructions/AGENTS.template.md` |\n\n## 프리셋\n\n자주 쓰는 도구/구성요소 조합을 프리셋으로 제공합니다.\n\n| 프리셋 ID | 포함 도구 | 설명 |\n|----------|----------|------|\n| `balanced-core` | Codex + Claude Code | 빠르게 시작하는 기본 구성 (추천) |\n| `full-suite` | Claude Code + Gemini CLI + Codex | 모든 구성요소 설치 |\n| `security-first` | Claude Code + Gemini CLI + Codex | 보안 점검/하드닝 중심 |\n| `frontend-refactor` | Claude Code + Codex | UI 컴포넌트화/성능 최적화 |\n| `prd-to-production` | Claude Code + Gemini CLI + Codex | PRD→구현 자동 파이프라인 |\n\n## CLI 옵션\n\n```\ntri-agent-manager [command] [options]\n\nCommands:\n  install          설치 (기본값)\n  update           업데이트\n  list             설치 가능 도구/구성요소 목록\n\nOptions:\n  --interactive    대화형 모드 (TUI 메뉴)\n  --preset <id>    프리셋으로 도구/구성요소 사전 선택\n  --tool <ids>     도구 지정 (쉼표 구분: claude-code,gemini,codex)\n  --components <ids>  구성요소 지정 (쉼표 구분: skills,commands,...)\n  --target <path>  설치 대상 경로 (기본: 현재 디렉토리)\n  --project-rules <policy>  규칙 파일 정책 (always|if-instructions|if-present|never)\n  --force          install 시 기존 파일 덮어쓰기\n  --dry-run        실제 파일 쓰기 없이 계획만 출력\n  --yes            확인 프롬프트 건너뛰기\n  --help           도움말\n```\n\n### install vs update\n\n| 명령 | 기존 파일 처리 | 용도 |\n|------|-------------|------|\n| `install` | 기존 파일 유지 (건너뜀) | 최초 설치, 추가 도구 설치 |\n| `install --force` | 기존 파일 덮어씀 | 템플릿 강제 재적용 |\n| `update` | 선택한 항목 업데이트 반영 | 버전 갱신, 유지보수 |\n\n## 도구별 구성요소\n\n<details>\n<summary><strong>Claude Code (.claude)</strong></summary>\n\n| 구성요소 ID | 설치 경로 | 설명 |\n|------------|----------|------|\n| `skills` | `.claude/skills` | 8개 운영 스킬 |\n| `subagents` | `.claude/agents` | 15개 역할별 서브에이전트 프롬프트 |\n| `agent-teams` | `.claude/agent-teams` | 리드+전문가 팀 구성 |\n| `commands` | `.claude/commands` | 20개 네이티브 슬래시 명령 |\n| `instructions` | `.claude/instructions` | `CLAUDE.md` 템플릿 |\n| `settings` | `.claude/settings` | 권한/보안 설정 템플릿 |\n\n규칙 파일: `<target>/CLAUDE.md`\n\n</details>\n\n<details>\n<summary><strong>Gemini CLI (.gemini)</strong></summary>\n\n| 구성요소 ID | 설치 경로 | 설명 |\n|------------|----------|------|\n| `skills` | `.gemini/skills` | 8개 운영 스킬 |\n| `agents` | `.gemini/agents` | 에이전트 역할 카탈로그 |\n| `artifacts` | `.gemini/artifacts` | 단계 입출력 아티팩트 계약 |\n| `workflows` | `.gemini/workflows` | 5개 워크플로우 + YAML 정의 |\n| `commands` | `.gemini/commands` | 20개 TOML 커스텀 명령 |\n| `instructions` | `.gemini/instructions` | `GEMINI.md` 템플릿 |\n| `settings` | `.gemini/settings` | 에디터 보안/권한 정책 |\n\n규칙 파일: `<target>/GEMINI.md`\n\n</details>\n\n<details>\n<summary><strong>Codex (.codex)</strong></summary>\n\n| 구성요소 ID | 설치 경로 | 설명 |\n|------------|----------|------|\n| `skills` | `.codex/skills` | 8개 운영 스킬 + 20개 명령 스킬 |\n| `instructions` | `.codex/instructions` | `AGENTS.md` 템플릿 + 권한 스니펫 |\n| `automations` | `.codex/automations` | 반복 작업 자동화 레시피 |\n| `commands` | `.codex/commands` | 공통 명령 계약 |\n| `settings` | `.codex/settings` | 런타임 권한 정책 |\n\n규칙 파일: `<target>/AGENTS.md`\n\n</details>\n\n## 개발자용\n\n### 파이프라인\n\n```\ncommon/* 수정 → npm run generate → npm run validate → npm publish\n```\n\n- `npm run generate` — 공통 정의를 플랫폼별 파일로 생성\n- `npm run generate:check` — 생성 결과가 최신인지 검증\n- `npm run validate` — 구조/계약/CLI 동작 검증\n- `npm run prepublish:check` — 배포 전 전체 검증\n\n### 규칙\n\n- 공통 원천은 `common/*`에서만 수정합니다.\n- `claude-code/*`, `codex/*`, `gemini/*`는 생성 결과로 취급합니다. 직접 수정하지 마세요.\n- 명령은 공통 계약에서 플랫폼별 네이티브 포맷으로 자동 변환됩니다:\n  - Claude Code: `commands/native/*.md` (Slash Command)\n  - Gemini CLI: `commands/definitions/*.toml` (Extension 커스텀 명령)\n  - Codex: `skills/cmd-*/SKILL.md` (스킬 변환)\n- 보안 정책(`common/settings/security-policy.json`)에서 플랫폼별 설정 템플릿으로 변환됩니다.\n\n### 검증 대상\n\n- 생성 결과 최신성\n- 스킬 frontmatter/절차 유효성 (CMD-*/WF-*/ART-* 참조 금지)\n- 명령/워크플로우 참조 무결성\n- 카탈로그/프리셋 무결성\n- Gemini Extension 구조 (`gemini-extension.json`, `.toml` 커스텀 명령)\n- CLI 스모크 테스트 (list/install/update + preset)\n\n## 라이선스\n\nMIT\n","readmeFilename":"README.md"}