[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AB7W2Z0PKFJC6ZR439EBAN":3,"skill-same:01M7AB7W2Z0PKFJC6ZR439EBAN":72},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M7AB7W2Z0PKFJC6ZR439EBAN","fleet",null,"Orchestrate bounded GPT-only Codex fleets across Sol, Terra, and Luna Max lanes. Use when a user asks for Fleet, parallel agents, deep multi-agent work, broad classification, independent verification, or a cron-friendly batch that must use ChatGPT-plan Codex authentication without third-party providers.","Babinci\u002Fcopernicus","Babinci","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F56366501?v=4","01M7AB6P4BSH34FQCJ8CVSF9J4","Wojciech Bieniek","01M7AB6P50AYTQF8F8JBS3QFQS",0,"MIT","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: fleet\ndescription: Orchestrate bounded GPT-only Codex fleets across Sol, Terra, and Luna Max lanes. Use when a user asks for Fleet, parallel agents, deep multi-agent work, broad classification, independent verification, or a cron-friendly batch that must use ChatGPT-plan Codex authentication without third-party providers.\n---\n\n# Copernicus Fleet\n\nRoute each seat to the cheapest GPT lane that can meet a checkable contract. The lead owns decomposition, width, evidence, and synthesis; workers never create more workers.\n\nResolve `\u003Cfleet-skill-dir>` to the directory containing this `SKILL.md`. Resolve\nall bundled scripts, references, and model policy files from that directory,\nnever from the user's current working directory.\n\nRead [guide.md](references\u002Fguide.md) when explaining Fleet, choosing between an\ninteractive mission and a batch, designing a first mission, or changing this\nskill. It describes the purpose, execution model, outputs, examples, privacy\nboundary, and limitations.\n\nOpen a fleet mission with:\n\n```text\nUsing Copernicus Fleet: GPT-only.\nMission: \u003Cone sentence>.\nExecute vs validate: \u003Cwhat runs now>.\n```\n\n## Route\n\nRead `fleet.yaml` before routing. It contains the GPT model bindings, named\npresets, and the active `default_preset`. The shipped default is `terra-first`:\nan omitted manifest lane resolves to Terra. Use `luna-breadth` when repeatable\nwidth should be the default, or name any lane explicitly to override a preset.\n\n- **Sol** (`gpt-5.6-sol`): ambiguous, high-value synthesis, architecture, and final verification. Use 3-4 seats only for an explicitly deep mission.\n- **Terra** (`gpt-5.6-terra`): default everyday implementation, grounded review, mapping, and reducers.\n- **Luna Max** (`gpt-5.6-luna`, `max`): clear, repeatable extraction, classification, and bounded shards. Run waves of at most five.\n\nRead [model-policy.md](references\u002Fmodel-policy.md) before changing lane names,\npresets, concurrency, model IDs, or reasoning efforts. If a configured model is\nunavailable to the signed-in account, report it as unavailable; never fall back\nto a non-GPT provider. For a different account catalog, copy `fleet.yaml`, bind\nonly observed GPT IDs, and pass the copy with `--config-file`.\n\n## Execute\n\n1. Run deterministic commands first. Do not spend a seat on a claim a test, parser, grep, or measurement can settle.\n2. Give every seat a unique role, bounded material, allowed paths, edit policy, typed output, evidence standard, falsifier, and stop rule. State `Do not spawn subagents`.\n3. Keep verification independent: an author does not certify its own result. Prefer a deterministic gate, then a fresh Terra or Sol review.\n4. Launch only independent seats in parallel. Use Luna waves for breadth, then deduplicate before synthesis.\n5. Treat timeout, quota, auth, and unavailable-model results as missing evidence, never as a vote.\n6. Verify every load-bearing claim against source or a runnable check before persisting it.\n\nUse Codex collaboration tools for interactive missions. For local scheduled or shell-driven batches, use the bundled runner:\n\n```bash\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py list\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py list --preset luna-breadth\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py batch manifest.jsonl --run-dir .copernicus\u002Fruns\u002Fexample --workdir . --dry-run\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py batch manifest.jsonl --run-dir .copernicus\u002Fruns\u002Fexample --workdir .\n```\n\nManifest rows contain `id`, `prompt`, and an optional `lane`. Omitting `lane`\nuses the active preset's default; an explicit lane always wins:\n\n```json\n{\"id\":\"review-api\",\"prompt\":\"prompts\u002Freview-api.md\"}\n{\"id\":\"map-api\",\"lane\":\"luna\",\"prompt\":\"prompts\u002Fmap-api.md\"}\n```\n\nThe runner is read-only by default, invokes `codex exec` without a shell, clamps concurrency to lane limits, stores one private answer\u002Fevents\u002Fstatus directory per seat, and returns nonzero if any seat fails. Use `--sandbox workspace-write` only for an explicitly authorized implementation batch. It never supports `danger-full-access`.\n\n## Seat contract\n\n```text\nROLE: \u003Cone bounded function>. Read-only unless exact writable paths follow. Do not spawn subagents.\nGOAL: \u003Cuser-visible destination>.\nSUCCESS: \u003Ccheckable result and verification>.\nMATERIAL: \u003Cexact paths or evidence slice>.\nCONSTRAINTS: GPT-only; no secrets; no destructive or external writes; cite evidence read.\nOUTPUT: \u003Cschema or headings>.\nFALSIFIER: \u003Cobservation that disproves the claim>.\nSTOP: emit UNVERIFIED when evidence is missing; stop after the bounded result.\n```\n\n## Hard rules\n\n- Use only model IDs beginning with `gpt-`; the runner enforces this again at execution.\n- Use the user's existing Codex sign-in. Never read, copy, print, or commit auth caches, API keys, or tokens.\n- Do not use provider proxies, OpenRouter, or third-party model fallbacks.\n- Do not let workers deploy, purchase, sign, publish, message people, or mutate external systems without the user's explicit authorization.\n- Do not create immortal agents. Durable state belongs in files, tests, receipts, or scheduled-task history.\n- Do not count correlated Luna outputs as independent validation.\n\nKeep seat provenance in the private run trace. In a normal conversational\nresponse, lead with the verified outcome and omit the roster. Show a compact\nroster only when the user requests execution provenance, or when a failed or\nunavailable seat or material dissent changes confidence. Then give verified\nclaims, rejected or unverified claims, and useful next commands.\n\nFleet is self-contained: interactive work uses Codex's collaboration surface;\nbatch work uses the bundled Python standard-library runner and the user's\nexisting `codex` executable. It needs no provider proxy, API-key router,\ndatabase, daemon, or private project files.\n","# Copernicus Fleet\n\nRoute each seat to the cheapest GPT lane that can meet a checkable contract. The lead owns decomposition, width, evidence, and synthesis; workers never create more workers.\n\nResolve `\u003Cfleet-skill-dir>` to the directory containing this `SKILL.md`. Resolve\nall bundled scripts, references, and model policy files from that directory,\nnever from the user's current working directory.\n\nRead [guide.md](references\u002Fguide.md) when explaining Fleet, choosing between an\ninteractive mission and a batch, designing a first mission, or changing this\nskill. It describes the purpose, execution model, outputs, examples, privacy\nboundary, and limitations.\n\nOpen a fleet mission with:\n\n```text\nUsing Copernicus Fleet: GPT-only.\nMission: \u003Cone sentence>.\nExecute vs validate: \u003Cwhat runs now>.\n```\n\n## Route\n\nRead `fleet.yaml` before routing. It contains the GPT model bindings, named\npresets, and the active `default_preset`. The shipped default is `terra-first`:\nan omitted manifest lane resolves to Terra. Use `luna-breadth` when repeatable\nwidth should be the default, or name any lane explicitly to override a preset.\n\n- **Sol** (`gpt-5.6-sol`): ambiguous, high-value synthesis, architecture, and final verification. Use 3-4 seats only for an explicitly deep mission.\n- **Terra** (`gpt-5.6-terra`): default everyday implementation, grounded review, mapping, and reducers.\n- **Luna Max** (`gpt-5.6-luna`, `max`): clear, repeatable extraction, classification, and bounded shards. Run waves of at most five.\n\nRead [model-policy.md](references\u002Fmodel-policy.md) before changing lane names,\npresets, concurrency, model IDs, or reasoning efforts. If a configured model is\nunavailable to the signed-in account, report it as unavailable; never fall back\nto a non-GPT provider. For a different account catalog, copy `fleet.yaml`, bind\nonly observed GPT IDs, and pass the copy with `--config-file`.\n\n## Execute\n\n1. Run deterministic commands first. Do not spend a seat on a claim a test, parser, grep, or measurement can settle.\n2. Give every seat a unique role, bounded material, allowed paths, edit policy, typed output, evidence standard, falsifier, and stop rule. State `Do not spawn subagents`.\n3. Keep verification independent: an author does not certify its own result. Prefer a deterministic gate, then a fresh Terra or Sol review.\n4. Launch only independent seats in parallel. Use Luna waves for breadth, then deduplicate before synthesis.\n5. Treat timeout, quota, auth, and unavailable-model results as missing evidence, never as a vote.\n6. Verify every load-bearing claim against source or a runnable check before persisting it.\n\nUse Codex collaboration tools for interactive missions. For local scheduled or shell-driven batches, use the bundled runner:\n\n```bash\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py list\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py list --preset luna-breadth\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py batch manifest.jsonl --run-dir .copernicus\u002Fruns\u002Fexample --workdir . --dry-run\npython3 \u003Cfleet-skill-dir>\u002Fscripts\u002Ffleet.py batch manifest.jsonl --run-dir .copernicus\u002Fruns\u002Fexample --workdir .\n```\n\nManifest rows contain `id`, `prompt`, and an optional `lane`. Omitting `lane`\nuses the active preset's default; an explicit lane always wins:\n\n```json\n{\"id\":\"review-api\",\"prompt\":\"prompts\u002Freview-api.md\"}\n{\"id\":\"map-api\",\"lane\":\"luna\",\"prompt\":\"prompts\u002Fmap-api.md\"}\n```\n\nThe runner is read-only by default, invokes `codex exec` without a shell, clamps concurrency to lane limits, stores one private answer\u002Fevents\u002Fstatus directory per seat, and returns nonzero if any seat fails. Use `--sandbox workspace-write` only for an explicitly authorized implementation batch. It never supports `danger-full-access`.\n\n## Seat contract\n\n```text\nROLE: \u003Cone bounded function>. Read-only unless exact writable paths follow. Do not spawn subagents.\nGOAL: \u003Cuser-visible destination>.\nSUCCESS: \u003Ccheckable result and verification>.\nMATERIAL: \u003Cexact paths or evidence slice>.\nCONSTRAINTS: GPT-only; no secrets; no destructive or external writes; cite evidence read.\nOUTPUT: \u003Cschema or headings>.\nFALSIFIER: \u003Cobservation that disproves the claim>.\nSTOP: emit UNVERIFIED when evidence is missing; stop after the bounded result.\n```\n\n## Hard rules\n\n- Use only model IDs beginning with `gpt-`; the runner enforces this again at execution.\n- Use the user's existing Codex sign-in. Never read, copy, print, or commit auth caches, API keys, or tokens.\n- Do not use provider proxies, OpenRouter, or third-party model fallbacks.\n- Do not let workers deploy, purchase, sign, publish, message people, or mutate external systems without the user's explicit authorization.\n- Do not create immortal agents. Durable state belongs in files, tests, receipts, or scheduled-task history.\n- Do not count correlated Luna outputs as independent validation.\n\nKeep seat provenance in the private run trace. In a normal conversational\nresponse, lead with the verified outcome and omit the roster. Show a compact\nroster only when the user requests execution provenance, or when a failed or\nunavailable seat or material dissent changes confidence. Then give verified\nclaims, rejected or unverified claims, and useful next commands.\n\nFleet is self-contained: interactive work uses Codex's collaboration surface;\nbatch work uses the bundled Python standard-library runner and the user's\nexisting `codex` executable. It needs no provider proxy, API-key router,\ndatabase, daemon, or private project files.\n",{"name":5,"description":7},"main","caa1cc6207632cbd529bd320d2e3818f3078c276","plugins\u002Fcopernicus\u002Fskills\u002Ffleet","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus","https:\u002F\u002Fgithub.com\u002FBabinci","2026-09-30 00:49:12",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add Babinci\u002Fcopernicus --skill fleet",{"tool":36,"command":37},"bunx","bunx skills add Babinci\u002Fcopernicus --skill fleet",{"tool":39,"command":40},"pnpm","pnpm dlx skills add Babinci\u002Fcopernicus --skill fleet","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002FSKILL.md",true,[44,50,54,58,62,68],{"path":45,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":48,"sourceHtmlUrl":49},"agents\u002Fopenai.yaml","REFERENCE","参考文档",225,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Fagents\u002Fopenai.yaml",{"path":51,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":52,"sourceHtmlUrl":53},"fleet.yaml",592,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Ffleet.yaml",{"path":55,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":56,"sourceHtmlUrl":57},"references\u002Fguide.md",6214,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Freferences\u002Fguide.md",{"path":59,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":60,"sourceHtmlUrl":61},"references\u002Fmodel-policy.md",2014,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Freferences\u002Fmodel-policy.md",{"path":63,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":66,"sourceHtmlUrl":67},"scripts\u002Ffleet.py","SCRIPT","脚本",13267,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Fscripts\u002Ffleet.py",{"path":69,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":70,"sourceHtmlUrl":71},"scripts\u002Ftest_fleet.py",7246,"https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fplugins\u002Fcopernicus\u002Fskills\u002Ffleet\u002Fscripts\u002Ftest_fleet.py",[73,80,87,94,101,108],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7AB9C5QZW0WQKQGT36KMRT7","upstream-dmg-watchdog","Run the versioned upstream DMG probe, repair worker, guarded PR, and sequential Nix refresh flow.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Fscripts\u002Fautomation\u002Fupstream-dmg-watchdog\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7AB9AW84M3TKGRCM7EX4ZN3","Unsafe Symlink","Use when testing symlink safety.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Funsafe-symlink\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M7AB99PC174BDQGPWKW0F6DZ","Mac Only","Use when automating Finder with AppleScript.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fplatform-macos\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":19},"01M7AB985WEE2Q5ZA0MG21V96T","Instruction Only","Use when answering with reusable written instructions.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Finstruction-only\u002FSKILL.md",[],[],{"skillNo":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"publishedAt":19},"01M7AB95F9KS3TQFG183BB95PB","Desktop Act","Use when clicking and typing in a Linux desktop app.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fdesktop-act\u002FSKILL.md",[],[],{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":19},"01M7AB93YB9J8BGZ9M28KB9CPN","Collision Tripwire","Use when testing script handling.","https:\u002F\u002Fgithub.com\u002FBabinci\u002Fcopernicus\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fcollision-tripwire\u002FSKILL.md",[],[]]