[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6PHBSYAXQV34PMHR2XM2FG9":3,"skill-same:01M6PHBSYAXQV34PMHR2XM2FG9":69},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":22,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":48,"commitSha":49,"skillDirectory":50,"forksCount":51,"repositoryHtmlUrl":52,"repositoryHomepageUrl":6,"ownerHtmlUrl":53,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":54,"install":55,"resources":68},"01M6PHBSYAXQV34PMHR2XM2FG9","scheduler-registration",null,"Use when the user asks to design or implement a scheduler, queue worker, automation runner, polling loop, batch processor, cron-like flow, job orchestrator, or start\u002Fstop\u002Frestart\u002Finspect controls for long-running work in ClaudePlay. Covers session-bound server-side async loop design, lifecycle rules, scheduler registration, metadata, inspect\u002Fstart\u002Fstop\u002Frestart APIs, and MCP observability. Do not use for simple one-shot panel actions or a single direct tool call.","DevelopmentDummy\u002Fclaude-play","DevelopmentDummy","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F269318128?v=4","01M6PH7WG74ND790ED9698KRBP","01M6PH7WM7KC9N0V51ZCQ43P8B",3,"MIT","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fcomfyui\u002Fskills\u002Fscheduler-registration\u002FSKILL.md",[17],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"scheduler registration","scheduler","registration","direct","loop design","inspect","batch processor","worker","runner","orchestrator","work","design","loop","processor","flow","start","actions","server","batch","MCP","2026-11-10 10:12:30",false,"---\nname: scheduler-registration\ndescription: Use when the user asks to design or implement a scheduler, queue worker, automation runner, polling loop, batch processor, cron-like flow, job orchestrator, or start\u002Fstop\u002Frestart\u002Finspect controls for long-running work in ClaudePlay. Covers session-bound server-side async loop design, lifecycle rules, scheduler registration, metadata, inspect\u002Fstart\u002Fstop\u002Frestart APIs, and MCP observability. Do not use for simple one-shot panel actions or a single direct tool call.\n---\n\n# Scheduler Registration\n\nUse this skill when a builder or service task introduces:\n\n- scheduler\n- scheduler registration\n- automation runner\n- automation pipeline\n- batch loop\n- batch processor\n- background-like polling\n- polling loop\n- queue worker\n- job runner\n- job orchestrator\n- cron-like flow\n- start\u002Fstop\u002Frestart controls\n- service status inspection\n\nDo not use this for simple panel buttons that only trigger one-shot actions.\n\nTypical trigger requests:\n\n- \"스케줄러 설계해줘\"\n- \"자동 실행 루프를 붙여줘\"\n- \"배치 러너를 서비스 엔진으로 옮기자\"\n- \"start\u002Fstop\u002Frestart API를 만들자\"\n- \"inspect\u002Fstatus MCP 도구를 붙이자\"\n- \"백그라운드처럼 도는 작업을 등록하자\"\n\n## Default Mental Model\n\nClaudePlay schedulers should default to:\n\n- service-side\n- session-bound\n- async loop based\n- tick driven\n- observable through API\u002FMCP\n\nUnless the service explicitly needs durable background execution, do not propose a separate worker process or \"background thread\".\n\nPrefer this wording:\n\n- \"server-side async loop\"\n- \"session-bound scheduler handle\"\n- \"tick-based runner\"\n\nAvoid this wording unless it is literally true:\n\n- \"background thread\"\n- \"independent daemon\"\n- \"always-on worker\"\n\n## Registration Rules\n\nWhen registering a scheduler, define all of these clearly:\n\n1. Ownership\n   - Which service module owns the handle?\n   - Where is runtime state stored?\n\n2. Lifecycle\n   - How it starts\n   - How it stops\n   - What happens on completion\n   - What happens on error\n   - What happens when connected clients become zero\n\n3. Tick contract\n   - What one tick does\n   - How the next action is chosen\n   - Sleep policy after work \u002F idle\n   - Stop flag check points\n\n4. Observability\n   - inspect API\u002FMCP route\n   - client\u002Fsession counts\n   - scheduler metadata\n   - last error and last tick time\n\n5. Controls\n   - inspect\n   - start\n   - stop\n   - restart\n\n## Recommended Metadata\n\nAlways include observable scheduler metadata where useful:\n\n- `label`\n- `source`\n- `requestedBy`\n- `note`\n- `phase`\n- `startedAt`\n- `lastTickAt`\n- `lastError`\n\nIf multiple schedulers may exist later, metadata is not optional in practice.\n\n## Safe Default Policy\n\nFor ClaudePlay, the default-safe scheduler policy is:\n\n- session-bound\n- stop when no clients remain in the session\n- stop on explicit user request\n- stop on completion\n- stop on fatal error\n\nOnly propose durable execution beyond session lifetime when the user explicitly wants a real job runner.\n\n## Recommended Architecture\n\nPrefer this shape:\n\n1. panel\u002Fbutton or MCP tool calls service API\n2. service API validates session activity\n3. runtime registry creates or reuses a scheduler handle\n4. handle runs an async loop\n5. each tick calls domain logic or session tool\n6. scheduler state is exposed through inspect\u002Fstatus APIs\n\n## Anti-Patterns\n\nAvoid these unless there is a strong reason:\n\n- browser-local timer loops as the primary scheduler\n- panel HTML owning long-running orchestration\n- hidden loops without inspectable status\n- start\u002Fstop without runtime metadata\n- session client count and scheduler state diverging silently\n\n## Output Expectations\n\nWhen asked to design or implement scheduler registration, produce:\n\n- runtime ownership point\n- lifecycle rules\n- start\u002Fstop\u002Frestart\u002Finspect surface\n- metadata schema\n- stop\u002Fcompletion policy\n- brief note on why this is session-bound or durable\n\nIf code changes are requested, keep the design consistent across:\n\n- service runtime\n- API routes\n- MCP tools\n- panel controls\n- status\u002Finspection views\n\n## Scheduler Notifications\n\n스케줄러의 `scheduler_tick` 반환값에 `notifications` 배열을 포함하면, 서버가 클라이언트(UI)와 AI 세션에 메시지를 자동으로 전달한다.\n\n### Notification Schema\n\n```javascript\n\u002F\u002F scheduler_tick 반환값 예시\nreturn {\n  success: true,\n  phase: \"source\",\n  did_work: true,\n  notifications: [\n    \u002F\u002F 클라이언트(UI)에 알림 — wsBroadcast로 전달\n    {\n      target: \"client\",\n      event: \"scheduler:progress\",     \u002F\u002F WebSocket 이벤트 타입 (기본: \"scheduler:notify\")\n      payload: { step: 3, total: 10 }  \u002F\u002F 자유 형태\n    },\n    \u002F\u002F AI에 이벤트 큐 — 다음 유저 메시지에 같이 전달\n    {\n      target: \"ai\",\n      mode: \"queue\",                   \u002F\u002F \"queue\" (기본) 또는 \"send\"\n      message: \"[SCHEDULER] 3\u002F10 완료\"\n    },\n    \u002F\u002F AI에 직접 턴 트리거 — 응답 중이면 완료 대기 후 전송\n    {\n      target: \"ai\",\n      mode: \"send\",\n      message: \"[SCHEDULER_COMPLETE] 작업이 완료되었습니다. 결과를 확인하고 유저에게 알려주세요.\"\n    }\n  ]\n};\n```\n\n### target별 동작\n\n| target | mode | 동작 |\n|--------|------|------|\n| `client` | — | `wsBroadcast(event, payload, { sessionId })` 호출. 해당 세션의 모든 클라이언트에 WebSocket 이벤트 전달 |\n| `ai` | `queue` (기본) | `instance.queueEvent(message)` 호출. 이벤트 큐에 쌓이고, 다음 유저 메시지 전송 시 AI에 함께 전달 |\n| `ai` | `send` | `instance.sendMessage(message)` 호출. AI에 직접 새 턴을 트리거. AI가 응답 중이면 완료 대기 후 전송 |\n\n### 사용 가이드\n\n- **진행 상황 알림**: `target: \"client\"` — 프론트엔드에서 토스트, 프로그레스 바, 패널 업데이트에 활용\n- **AI에 중간 보고**: `target: \"ai\", mode: \"queue\"` — 급하지 않은 정보. 다음 대화 턴에 자연스럽게 포함\n- **AI에 즉시 알림**: `target: \"ai\", mode: \"send\"` — 완료\u002F에러 등 AI가 즉시 반응해야 하는 이벤트. 대화가 끊긴 상태에서도 AI 턴을 시작함\n- 매 틱마다 0~N개 알림 가능. 완료 틱(`completed: true`)에도 알림이 처리된 후 루프가 종료됨\n- notification 처리 중 에러는 로그만 남기고 스케줄러 루프는 중단하지 않음\n","# Scheduler Registration\n\nUse this skill when a builder or service task introduces:\n\n- scheduler\n- scheduler registration\n- automation runner\n- automation pipeline\n- batch loop\n- batch processor\n- background-like polling\n- polling loop\n- queue worker\n- job runner\n- job orchestrator\n- cron-like flow\n- start\u002Fstop\u002Frestart controls\n- service status inspection\n\nDo not use this for simple panel buttons that only trigger one-shot actions.\n\nTypical trigger requests:\n\n- \"스케줄러 설계해줘\"\n- \"자동 실행 루프를 붙여줘\"\n- \"배치 러너를 서비스 엔진으로 옮기자\"\n- \"start\u002Fstop\u002Frestart API를 만들자\"\n- \"inspect\u002Fstatus MCP 도구를 붙이자\"\n- \"백그라운드처럼 도는 작업을 등록하자\"\n\n## Default Mental Model\n\nClaudePlay schedulers should default to:\n\n- service-side\n- session-bound\n- async loop based\n- tick driven\n- observable through API\u002FMCP\n\nUnless the service explicitly needs durable background execution, do not propose a separate worker process or \"background thread\".\n\nPrefer this wording:\n\n- \"server-side async loop\"\n- \"session-bound scheduler handle\"\n- \"tick-based runner\"\n\nAvoid this wording unless it is literally true:\n\n- \"background thread\"\n- \"independent daemon\"\n- \"always-on worker\"\n\n## Registration Rules\n\nWhen registering a scheduler, define all of these clearly:\n\n1. Ownership\n   - Which service module owns the handle?\n   - Where is runtime state stored?\n\n2. Lifecycle\n   - How it starts\n   - How it stops\n   - What happens on completion\n   - What happens on error\n   - What happens when connected clients become zero\n\n3. Tick contract\n   - What one tick does\n   - How the next action is chosen\n   - Sleep policy after work \u002F idle\n   - Stop flag check points\n\n4. Observability\n   - inspect API\u002FMCP route\n   - client\u002Fsession counts\n   - scheduler metadata\n   - last error and last tick time\n\n5. Controls\n   - inspect\n   - start\n   - stop\n   - restart\n\n## Recommended Metadata\n\nAlways include observable scheduler metadata where useful:\n\n- `label`\n- `source`\n- `requestedBy`\n- `note`\n- `phase`\n- `startedAt`\n- `lastTickAt`\n- `lastError`\n\nIf multiple schedulers may exist later, metadata is not optional in practice.\n\n## Safe Default Policy\n\nFor ClaudePlay, the default-safe scheduler policy is:\n\n- session-bound\n- stop when no clients remain in the session\n- stop on explicit user request\n- stop on completion\n- stop on fatal error\n\nOnly propose durable execution beyond session lifetime when the user explicitly wants a real job runner.\n\n## Recommended Architecture\n\nPrefer this shape:\n\n1. panel\u002Fbutton or MCP tool calls service API\n2. service API validates session activity\n3. runtime registry creates or reuses a scheduler handle\n4. handle runs an async loop\n5. each tick calls domain logic or session tool\n6. scheduler state is exposed through inspect\u002Fstatus APIs\n\n## Anti-Patterns\n\nAvoid these unless there is a strong reason:\n\n- browser-local timer loops as the primary scheduler\n- panel HTML owning long-running orchestration\n- hidden loops without inspectable status\n- start\u002Fstop without runtime metadata\n- session client count and scheduler state diverging silently\n\n## Output Expectations\n\nWhen asked to design or implement scheduler registration, produce:\n\n- runtime ownership point\n- lifecycle rules\n- start\u002Fstop\u002Frestart\u002Finspect surface\n- metadata schema\n- stop\u002Fcompletion policy\n- brief note on why this is session-bound or durable\n\nIf code changes are requested, keep the design consistent across:\n\n- service runtime\n- API routes\n- MCP tools\n- panel controls\n- status\u002Finspection views\n\n## Scheduler Notifications\n\n스케줄러의 `scheduler_tick` 반환값에 `notifications` 배열을 포함하면, 서버가 클라이언트(UI)와 AI 세션에 메시지를 자동으로 전달한다.\n\n### Notification Schema\n\n```javascript\n\u002F\u002F scheduler_tick 반환값 예시\nreturn {\n  success: true,\n  phase: \"source\",\n  did_work: true,\n  notifications: [\n    \u002F\u002F 클라이언트(UI)에 알림 — wsBroadcast로 전달\n    {\n      target: \"client\",\n      event: \"scheduler:progress\",     \u002F\u002F WebSocket 이벤트 타입 (기본: \"scheduler:notify\")\n      payload: { step: 3, total: 10 }  \u002F\u002F 자유 형태\n    },\n    \u002F\u002F AI에 이벤트 큐 — 다음 유저 메시지에 같이 전달\n    {\n      target: \"ai\",\n      mode: \"queue\",                   \u002F\u002F \"queue\" (기본) 또는 \"send\"\n      message: \"[SCHEDULER] 3\u002F10 완료\"\n    },\n    \u002F\u002F AI에 직접 턴 트리거 — 응답 중이면 완료 대기 후 전송\n    {\n      target: \"ai\",\n      mode: \"send\",\n      message: \"[SCHEDULER_COMPLETE] 작업이 완료되었습니다. 결과를 확인하고 유저에게 알려주세요.\"\n    }\n  ]\n};\n```\n\n### target별 동작\n\n| target | mode | 동작 |\n|--------|------|------|\n| `client` | — | `wsBroadcast(event, payload, { sessionId })` 호출. 해당 세션의 모든 클라이언트에 WebSocket 이벤트 전달 |\n| `ai` | `queue` (기본) | `instance.queueEvent(message)` 호출. 이벤트 큐에 쌓이고, 다음 유저 메시지 전송 시 AI에 함께 전달 |\n| `ai` | `send` | `instance.sendMessage(message)` 호출. AI에 직접 새 턴을 트리거. AI가 응답 중이면 완료 대기 후 전송 |\n\n### 사용 가이드\n\n- **진행 상황 알림**: `target: \"client\"` — 프론트엔드에서 토스트, 프로그레스 바, 패널 업데이트에 활용\n- **AI에 중간 보고**: `target: \"ai\", mode: \"queue\"` — 급하지 않은 정보. 다음 대화 턴에 자연스럽게 포함\n- **AI에 즉시 알림**: `target: \"ai\", mode: \"send\"` — 완료\u002F에러 등 AI가 즉시 반응해야 하는 이벤트. 대화가 끊긴 상태에서도 AI 턴을 시작함\n- 매 틱마다 0~N개 알림 가능. 완료 틱(`completed: true`)에도 알림이 처리된 후 루프가 종료됨\n- notification 처리 중 에러는 로그만 남기고 스케줄러 루프는 중단하지 않음\n",{"name":5,"description":7},"main","87c4e526fb2dfb2c6bfeeb9f901d6320fcb457ec","data\u002Ftools\u002Fcomfyui\u002Fskills\u002Fscheduler-registration",0,"https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy","2026-10-07 22:33:19",{"source":8,"skillName":5,"commands":56,"prompt":66,"zipAvailable":67},[57,60,63],{"tool":58,"command":59},"npx","npx skills add DevelopmentDummy\u002Fclaude-play --skill scheduler-registration",{"tool":61,"command":62},"bunx","bunx skills add DevelopmentDummy\u002Fclaude-play --skill scheduler-registration",{"tool":64,"command":65},"pnpm","pnpm dlx skills add DevelopmentDummy\u002Fclaude-play --skill scheduler-registration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fcomfyui\u002Fskills\u002Fscheduler-registration\u002FSKILL.md",true,[],[70,90,104,123,133,142],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":80,"publishedAt":43},"01M6PHD8Y8CNRBG3FR2ZAHWX5J","generate-video","claude-play-bridge MCP 서버로 영상을 생성한다. 로컬 ComfyUI의 영상 워크플로 패키지(MiniMax H3 = 영상+오디오 동시 생성, WAN 2.2 i2v, Z-Image→영상)를 호출한다. 렌더가 수 분~수십 분 걸리므로 async 큐잉 + 파일 폴링이 기본 절차다. \"영상\", \"동영상\", \"비디오\", \"video\", \"i2v\", \"t2v\", \"클립 만들어\", \"움직이는\", \"애니메이션 영상\" 요청에 트리거된다.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fscripts\u002Fexternal-package\u002Fskills\u002Fgenerate-video\u002FSKILL.md",[76,79],{"categoryNo":77,"categoryName":78,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000016","图像与多媒体",{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[81,82,83,42,84,85,86,87,88,89],"Video","generate","generate video","image","play","bridge","minimax","async","minimax h3",{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":98,"publishedAt":43},"01M6PHD3MAQ9QYQQFG50WBP5JZ","generate-image","claude-play-bridge MCP 서버로 이미지를 생성한다. 기본 백엔드는 ComfyUI(로컬 GPU, 워크플로 패키지 기반)이고, 사용자가 명시적으로 요청할 때만 GPT\u002FGemini 백엔드를 사용한다.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fscripts\u002Fexternal-package\u002Fskills\u002Fgenerate-image\u002FSKILL.md",[96,97],{"categoryNo":77,"categoryName":78,"taxonomyType":20,"taxonomyTypeDesc":21},{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[99,82,84,86,42,100,85,101,102,103],"generate image","gemini","gpu","gpt","comfyui",{"skillNo":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":111,"publishedAt":43},"01M6PHD0134XK1NQNK4YMZB2E1","generate-image-openai","OpenAI\u002FGPT 이미지 생성(GPT Image 품질 \u002F 텍스트 렌더링·이미지 편집 강점). 기본 백엔드는 Codex CLI 구독(건당 과금 없음). ⚠️ 사용자가 이 백엔드를 명시적으로 지정했을 때만 사용한다(예 \"GPT로\u002FOpenAI로 그려줘\"). 기본 이미지 생성은 generate-image(ComfyUI)이며, 사용자가 콕 집어 요청하지 않았다면 절대 OpenAI를 자동으로 선택하지 마라.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fopenai\u002Fskills\u002Fgenerate-image-openai\u002FSKILL.md",[110],{"categoryNo":77,"categoryName":78,"taxonomyType":20,"taxonomyTypeDesc":21},[112,99,113,82,84,114,115,116,117,118,102,119,120,121,103,122],"generate image openai","openai","렌더링","cli","이미지","codex","codex cli","Bash","read","write","텍스트",{"skillNo":124,"name":125,"nameZh":6,"description":126,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":127,"categories":128,"keywords":129,"publishedAt":43},"01M6PHCTMBPTG56KMK8VH5SCTE","generate-spritesheet","Gemini 이미지 생성으로 게임용 아이콘 스프라이트시트를 만들고 자동 슬라이스한다. 아이콘, UI 리소스, 뱃지, 마커 등 여러 개의 동일 카테고리 에셋이 필요할 때 사용한다.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fgemini\u002Fskills\u002Fgenerate-spritesheet\u002FSKILL.md",[],[130,82,100,131,119,120,121,132,116],"generate spritesheet","ui","edit",{"skillNo":134,"name":135,"nameZh":6,"description":136,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":137,"categories":138,"keywords":140,"publishedAt":43},"01M6PHCPR07KYS2RPNZER9WS2S","generate-image-gemini","Gemini API로 이미지를 생성한다. 패널 배경, UI 리소스, 아이콘, 장면 삽화 등 디자인 리소스 제작에 사용한다. ComfyUI가 없을 때의 이미지 생성 대안이기도 하다.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fgemini\u002Fskills\u002Fgenerate-image-gemini\u002FSKILL.md",[139],{"categoryNo":77,"categoryName":78,"taxonomyType":20,"taxonomyTypeDesc":21},[141,100,84,82,99,121,120,119,131,116],"generate image gemini",{"skillNo":143,"name":144,"nameZh":6,"description":145,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":146,"categories":147,"keywords":152,"publishedAt":43},"01M6PHCJD7S7KT64BB9F215TP6","workflow-research","새 ComfyUI 워크플로우를 설계하거나 신규 모델, 양자화 모델, LoRA 개발 전략을 비교할 때 분석 틀을 잡는다.","https:\u002F\u002Fgithub.com\u002FDevelopmentDummy\u002Fclaude-play\u002Fblob\u002Fmain\u002Fdata\u002Ftools\u002Fcomfyui\u002Fskills\u002Fworkflow-research\u002FSKILL.md",[148,151],{"categoryNo":149,"categoryName":150,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000021","资料检索与研究",{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[153,154,155,103,120,156],"workflow","research","workflow research","LoRA"]