[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-93048":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":9,"languages":9,"totalLinesOfCode":9,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":13,"stars7d":14,"stars30d":15,"stars90d":13,"forks30d":13,"starsTrendScore":13,"compositeScore":16,"rankGlobal":9,"rankLanguage":9,"license":17,"archived":18,"fork":18,"defaultBranch":19,"hasWiki":20,"hasPages":18,"topics":21,"createdAt":9,"pushedAt":9,"updatedAt":22,"readmeContent":23,"aiSummary":24,"trendingCount":13,"starSnapshotCount":13,"syncStatus":25,"lastSyncTime":26,"discoverSource":27},93048,"gpt-5-6-relay","Forward-Future\u002Fgpt-5-6-relay","Forward-Future","A public Codex skill for routing work through GPT-5.6 Sol, Terra, and Luna child threads.",null,146,21,4,0,1,44,48.93,"MIT License",false,"main",true,[],"2026-07-22 04:02:08","# GPT-5.6 Relay\n\nA public Codex skill for routing one task through model-specific GPT-5.6 child threads.\n\nThe relay gives each phase to the model and reasoning effort that fit it best, while preserving visible thread history and concrete handoffs:\n\n```text\nSol coordinator\n    -> Terra implementation\n    -> Sol review\n    -> Luna release\n```\n\nThe route adapts to the task. Small mechanical work may use only Luna. Ambiguous architecture may stay with Sol longer. Deployment is always a separate, authorized Luna phase.\n\n## Install\n\nCopy `.agents\u002Fskills\u002Fgpt-5-6-relay\u002F` into a project that supports Codex project skills. Then invoke it explicitly:\n\n```text\nInvoke \u002Fgpt-5-6-relay with this task: \u003Ctask>\n```\n\nThe skill uses Codex app thread tools to create visible child threads with explicit model and reasoning settings. It does not use hidden subprocesses.\n\n## Model routing\n\n| Model | Best fit | Relay starting effort |\n| --- | --- | --- |\n| Sol | Planning, architecture, ambiguity, hard diagnosis | Extra High |\n| Terra | Implementation, tests, bounded debugging | High |\n| Luna | Reconnaissance, mechanical checks, release execution | Light or Medium |\n\nThe user-facing effort levels are Light, Medium, High, Extra High, and Ultra. Sol and Terra support all five; Luna supports all except Ultra. Ultra enables automatic delegation on Sol and Terra, so the relay normally stops at Extra High and uses Ultra only when nested delegation is intentional.\n\n## What the skill guarantees\n\n- Each child thread has one role, one deliverable, and one checkable gate.\n- A dependent child is not started until its required artifact exists.\n- Corrections stay in the same child thread; changing model responsibility creates a new child.\n- Only one thread writes to a checkout at a time.\n- Deployment follows the repository's clean integrated-main rules in `DEPLOYMENT.md`.\n- The final report names the actual thread IDs, models, efforts, artifacts, checks, and deployment evidence.\n\n## Requirements\n\nThe host must expose Codex app tools for listing projects, creating threads, reading threads, and sending background messages with model and effort controls. If those controls are unavailable, the skill reports the proposed route instead of pretending that model routing happened.\n\n## License\n\nMIT. See [LICENSE](LICENSE).\n","这是一个面向Codex平台的模型路由技能，用于将单个任务按阶段分发至GPT-5.6系列专用子线程（Sol\u002FTerra\u002FLuna），实现规划、实现与发布环节的职责分离与显式协同。核心特点是基于任务复杂度动态适配模型角色与推理强度（Light–Extra High），全程保留可追溯的线程历史、明确交付物与校验门禁，并确保部署符合集成主干规范。适用于需要结构化AI协作流程的软件工程自动化场景，如需求分析、代码生成、测试验证与安全发布等端到端开发任务。",2,"2026-07-11 02:30:44","CREATED_QUERY"]