[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7B3G8GVRCSWS6XR2PZ5JBCY":3,"skill-same:01M7B3G8GVRCSWS6XR2PZ5JBCY":45},{"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":27,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":29,"updatedAt":30,"install":31,"resources":44},"01M7B3G8GVRCSWS6XR2PZ5JBCY","choosing-subagent-models",null,"Use when about to launch any subagent — a research or exploration agent, a reviewer, an implementer, a parallel fan-out, or a fork.","vadimvitvickiy\u002Fvv-agent-kit","vadimvitvickiy","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5019512?v=4","01M7B3ABRYF7KHBCDW6GB2XH0C","Vadim Vitvitskii","01M7B3ABSZYG9MXP1CC7BQYM25",0,"MIT","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fchoosing-subagent-models\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: choosing-subagent-models\ndescription: Use when about to launch any subagent — a research or exploration agent, a reviewer, an implementer, a parallel fan-out, or a fork.\n---\n\n# Choosing subagent models\n\n**Every spawn names its model.** An omitted `model` inherits the session's: a lookup from an Opus\nsession runs at full price, a final review from a cheap session runs cheap, and neither says so.\n\n- The kit's spawn hook refuses a launch with no `model`. It never picks one — pick here, by the job,\n  not by the model the session happens to run.\n- Use the aliases `haiku`, `sonnet`, `opus`, not version IDs. They survive a model release.\n\n## The table\n\n| The subagent's job | Model |\n|-|-|\n| Mechanical and fully specified — read a known file and extract a value, run a known test, build or lint command and report its output, summarize a log, apply a change whose exact text is given — where one command proves it done | `haiku` |\n| Gathering — web research, codebase exploration, \"where is X\", reading code to answer a question, implementing a plan task whose code is written out, reviewing one task's diff | `sonnet` |\n| Judgment — design and architecture, debugging after a fix has already failed, security review, the final review of a branch, implementing from prose where decisions are still open | `opus` |\n\n## Where the lines fall\n\n- **Haiku gets no search.** Unclear where to look, or what the thing is called, makes it research —\n  `sonnet`. A file haiku fails to find comes back as a confidently incomplete answer.\n- **Haiku stops at its first failure** — a failed command, an edit that does not apply, unexpected\n  output. Re-spawn on `sonnet`; do not retry. Cheap models take 2–3× the turns on multi-step work\n  and cost more overall. A red run from `vvkit:test-runner` is its result, not its failure.\n- **Sonnet is the floor for anything that decides correctness**: every reviewer, every implementer\n  that must make a test pass.\n- **Research fleets stay on `sonnet`**, even for an architectural question. Readers gather; the\n  judgment happens in the main conversation.\n- **Between two tiers, take the higher** and say so. When the higher one is a real cost — a fleet,\n  a long run — ask once, naming both tiers and what the higher one buys.\n- **Escalate, do not repeat.** A blocked implementer or a fix loop that is not converging goes up a\n  tier. The same spawn on the same model gives the same result.\n- **Two failed rounds end re-dispatching.** Take the core back into the main conversation, or send a\n  narrower piece one tier up. A wide brief that failed twice fails on scope, and a stronger model\n  inherits the same ambiguity.\n\n## Special cases\n\n- **A fork** runs on the parent's model and ignores `model`. Fork only when the inherited context is\n  worth the session's price.\n- **A named agent whose definition sets `model`** carries its author's default for an ordinary job.\n  Pass nothing when the table agrees; pass a tier up when the job is bigger — a whole branch,\n  security-sensitive code. Never a lower one.\n- **The user named a model.** Theirs wins.\n- **`CLAUDE_CODE_SUBAGENT_MODEL`** forces one model onto every subagent. A debugging tool, not a\n  policy.\n- `KIT_SUBAGENT_MODEL_GATE=0` turns the spawn hook off. The table still applies.\n","# Choosing subagent models\n\n**Every spawn names its model.** An omitted `model` inherits the session's: a lookup from an Opus\nsession runs at full price, a final review from a cheap session runs cheap, and neither says so.\n\n- The kit's spawn hook refuses a launch with no `model`. It never picks one — pick here, by the job,\n  not by the model the session happens to run.\n- Use the aliases `haiku`, `sonnet`, `opus`, not version IDs. They survive a model release.\n\n## The table\n\n| The subagent's job | Model |\n|-|-|\n| Mechanical and fully specified — read a known file and extract a value, run a known test, build or lint command and report its output, summarize a log, apply a change whose exact text is given — where one command proves it done | `haiku` |\n| Gathering — web research, codebase exploration, \"where is X\", reading code to answer a question, implementing a plan task whose code is written out, reviewing one task's diff | `sonnet` |\n| Judgment — design and architecture, debugging after a fix has already failed, security review, the final review of a branch, implementing from prose where decisions are still open | `opus` |\n\n## Where the lines fall\n\n- **Haiku gets no search.** Unclear where to look, or what the thing is called, makes it research —\n  `sonnet`. A file haiku fails to find comes back as a confidently incomplete answer.\n- **Haiku stops at its first failure** — a failed command, an edit that does not apply, unexpected\n  output. Re-spawn on `sonnet`; do not retry. Cheap models take 2–3× the turns on multi-step work\n  and cost more overall. A red run from `vvkit:test-runner` is its result, not its failure.\n- **Sonnet is the floor for anything that decides correctness**: every reviewer, every implementer\n  that must make a test pass.\n- **Research fleets stay on `sonnet`**, even for an architectural question. Readers gather; the\n  judgment happens in the main conversation.\n- **Between two tiers, take the higher** and say so. When the higher one is a real cost — a fleet,\n  a long run — ask once, naming both tiers and what the higher one buys.\n- **Escalate, do not repeat.** A blocked implementer or a fix loop that is not converging goes up a\n  tier. The same spawn on the same model gives the same result.\n- **Two failed rounds end re-dispatching.** Take the core back into the main conversation, or send a\n  narrower piece one tier up. A wide brief that failed twice fails on scope, and a stronger model\n  inherits the same ambiguity.\n\n## Special cases\n\n- **A fork** runs on the parent's model and ignores `model`. Fork only when the inherited context is\n  worth the session's price.\n- **A named agent whose definition sets `model`** carries its author's default for an ordinary job.\n  Pass nothing when the table agrees; pass a tier up when the job is bigger — a whole branch,\n  security-sensitive code. Never a lower one.\n- **The user named a model.** Theirs wins.\n- **`CLAUDE_CODE_SUBAGENT_MODEL`** forces one model onto every subagent. A debugging tool, not a\n  policy.\n- `KIT_SUBAGENT_MODEL_GATE=0` turns the spawn hook off. The table still applies.\n",{"name":5,"description":7},"main","cc22a222a76add5759a2df5fadf231a758691db5","skills\u002Fchoosing-subagent-models","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy","Senior iOS engineer. Architecture, SDKs and apps for iOS, tvOS and watchOS.","2026-11-13 19:10:29",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add vadimvitvickiy\u002Fvv-agent-kit --skill choosing-subagent-models",{"tool":37,"command":38},"bunx","bunx skills add vadimvitvickiy\u002Fvv-agent-kit --skill choosing-subagent-models",{"tool":40,"command":41},"pnpm","pnpm dlx skills add vadimvitvickiy\u002Fvv-agent-kit --skill choosing-subagent-models","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fchoosing-subagent-models\u002FSKILL.md",true,[],[46,53,60,67,74,81],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M7B3JN7VG0YJ4JB8RP6HYZSJ","no-heading","Use when proving that the validator rejects a body with no H1 and a force-loading @-link.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fno-heading\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M7B3JJMEW4E710ZG6HYJ7BWV","wrong-name-does-not-match-dir","Dispatches a subagent per task and reviews between each one.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fbroken-skill\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M7B3JGE9MFT1HRP41YRVKN97","writing-tests","Use when adding or changing behavior, fixing a bug, or judging whether an existing or generated test earns its place.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-tests\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7B3JD5X32VEBDBN7HB8E17Z","writing-skills","Use when authoring or editing a skill, agent, command or hook for this kit, or when deciding whether a piece of guidance belongs in a skill at all.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-skills\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7B3J992MZH2QVZRED0DRPD9","writing-project-instructions","Use when creating or editing a CLAUDE.md or AGENTS.md, when it has grown long, or when a rule in it also exists in an installed skill.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-project-instructions\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M7B3J55C5WPWN20QXP9ET848","writing-logs","Use when adding or reviewing log statements, or when deciding what level a message belongs at.","https:\u002F\u002Fgithub.com\u002Fvadimvitvickiy\u002Fvv-agent-kit\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-logs\u002FSKILL.md",[],[]]