[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6DHAP9JN9X7ZSWCWYTWN1KT":3,"skill-same:01M6DHAP9JN9X7ZSWCWYTWN1KT":79},{"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":17,"publishedAt":36,"linkOnly":37,"markdownContent":38,"markdownBody":39,"frontMatter":40,"refName":42,"commitSha":43,"skillDirectory":44,"forksCount":13,"repositoryHtmlUrl":45,"repositoryHomepageUrl":6,"ownerHtmlUrl":46,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":47,"install":48,"resources":60},"01M6DHAP9JN9X7ZSWCWYTWN1KT","team-lead",null,"Coordinate an Agent Team for parallelizable work — implementation, audit, research, or testing. Teaches the mechanical constraints (WAIT pattern, file ownership, model tiers) and lets you design the team. For orchestration_tier=cluster \u002F multi-agent runs.","JAuto-Corp\u002Fjustinventit","JAuto-Corp","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F254009312?v=4","01M6DH85D43FJAXD8RPBY6A91C","01M6DH85M60CDJBZ9W2CAFCPXE",0,"MIT","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fteam-lead\u002FSKILL.md",[],[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35],"team lead","team","multi-agent","wait","mechanical","pattern","agent team","coordinate","implementation","work","testing","model","constraints","audit","file","orchestration","research","design","2026-11-10 10:12:30",false,"---\nname: team-lead\ndescription: \"Coordinate an Agent Team for parallelizable work — implementation, audit, research, or testing. Teaches the mechanical constraints (WAIT pattern, file ownership, model tiers) and lets you design the team. For orchestration_tier=cluster \u002F multi-agent runs.\"\nuser-invocable: true\n---\n\n# Team Lead\n\n> Teammates are full Claude Code sessions. Coordinate them — don't inject content they already have.\n\nUse this skill when work is genuinely **parallelizable** — several independent units different agents can tackle at once. This is the multi-agent tier, relevant when the project runs a team (`orchestration_tier=cluster`). If the work is sequential or trivial, do it yourself.\n\nGood team work: a phase with independent frontend\u002Fbackend\u002Fdata items; an audit with one agent per concern; validation across many scenarios; a research question attacked from several angles.\n\n---\n\n## Lifecycle\n\n```\n1. ASSESS   what can actually run in parallel\n2. CREATE   TeamCreate(team_name: \"descriptive-name\")\n3. SPAWN    teammates with self-contained prompts (see spawn-templates.md)\n4. TASK     create tasks with clear ownership + acceptance criteria\n5. ASSIGN   tasks to teammates\n6. WAIT     stop making tool calls — enter the WAIT state (below)\n7. PROCESS  handle messages as they arrive — checkpoint, reassign, unblock\n8. COMPLETE shut down teammates, TeamDelete\n```\n\nA framework, not a rigid script — step 7 may loop many times or once. Adapt.\n\n---\n\n## The WAIT Pattern (non-negotiable)\n\nThe single most important rule. Get it wrong and the team is useless.\n\n**Messages arrive only when the lead is idle** — not making tool calls. During an active turn they queue indefinitely.\n\n```\nAfter spawning + assigning:\n1. Finish ALL tool calls\n2. Emit a brief text summary and STOP\n3. Teammate messages arrive as new turns\n4. Process one — then STOP again for the next\n```\n\nThe rhythm is **act → stop → receive → act → stop**. Make tool calls while \"waiting\" and you will never receive teammate messages.\n\n---\n\n## File Ownership (when teammates write code)\n\n**No two teammates may edit the same file** — there is no merge resolution between concurrent teammates. Before spawning, write an ownership manifest: each teammate gets a disjoint set of paths. If a task crosses zones, split it into two tasks with a dependency.\n\nFor read-only work (audit, research), ownership doesn't apply — many agents can read the same files safely.\n\n---\n\n## Model Tiers\n\nMatch the model to the cognitive load:\n\n| Tier | Use for |\n|-|-|\n| **opus** | Writing code, reviewing\u002Fauditing, planning — anything needing judgment or correctness reasoning |\n| **sonnet** | Exploration and context-gathering — near-opus accuracy at lower latency |\n| **haiku** | Mechanical coverage only — rote navigation, trivial lookups where being wrong is cheap |\n\nCatching one subtle bug pays for an opus reviewer many times over; don't cut the model on judgment work.\n\n---\n\n## Designing the Team\n\n- **Teammates have everything you have** — CLAUDE.md, skills, tools, git. Don't paste skill content into prompts; tell them which skill to invoke. Don't extract patterns; they can explore themselves.\n- **Name by role, not number** — `frontend-coder`, `schema-auditor`, `scenario-runner` — so tasks and messages stay legible.\n- **Task descriptions are the main lever** — specific files, acceptance criteria, and ownership boundaries produce good work; vague tasks produce vague work.\n- **Absolute paths everywhere** — a teammate's cwd is not the repo root.\n- Start smaller than you think: 3-5 focused teammates beat five scattered ones.\n\nSee `spawn-templates.md` for ready-made role skeletons — use, modify, or replace them.\n\n### Data-dependent teams\n\nWhen teammates share data (E2E, integration), the lead prepares it **before** spawning:\n1. Ensure the shared data\u002Ffixtures exist — don't assume.\n2. Smoke-test ONE unit before deploying the full team; fix a BLOCKED result before wasting more agents on it.\n3. One unit of work per task — agents report back, the lead decides what's next. Don't batch.\n4. Treat BLOCKED (can't even start) as a systemic signal — pause new assignments and investigate.\n\n---\n\n## Checkpoint & Recover\n\nCheckpoint after each report: `git log --oneline` for landed commits, run the appropriate validation, update `PROGRESS.md`\u002F`WORKING.md` for phase work.\n\nAfter a context reset or crash:\n1. Read the state chain (`WORKING.md`, `PROGRESS.md`).\n2. Check whether the team still exists; resume its `TaskList`, or `TeamDelete` and rebuild from remaining work.\n3. `git log` for commits since the last checkpoint — anything committed is done.\n\nCommitted work and `PROGRESS.md` are the durable checkpoints; in-flight tasks are not.\n\n---\n\n## Anti-Patterns\n\n| Don't | Why | Instead |\n|-|-|-|\n| Tool-call while waiting | Messages queue forever | STOP after assigning |\n| \"Let me just check…\" after spawning | Same blocking problem | Spawn → assign → text → STOP |\n| Two teammates on one file | No merge resolution | Split into dependent tasks |\n| Paste skill content into prompts | Teammates have skills natively | \"Invoke skill X first\" |\n| Poll TaskList in normal flow | Wasteful; messaging works when idle | TaskList only for recovery |\n| Disposable subagent for coordinated work | Subagents can't message back | Use an Agent Team |\n| Batch many units into one task | Context bloat, no incremental feedback | One unit per task |\n\n---\n\n## Related\n\n- Spawn skeletons: `spawn-templates.md`\n- Work lifecycle: the `work` skill\n- Explorer pattern & signal vocabulary: the `patterns` skill\n","# Team Lead\n\n> Teammates are full Claude Code sessions. Coordinate them — don't inject content they already have.\n\nUse this skill when work is genuinely **parallelizable** — several independent units different agents can tackle at once. This is the multi-agent tier, relevant when the project runs a team (`orchestration_tier=cluster`). If the work is sequential or trivial, do it yourself.\n\nGood team work: a phase with independent frontend\u002Fbackend\u002Fdata items; an audit with one agent per concern; validation across many scenarios; a research question attacked from several angles.\n\n---\n\n## Lifecycle\n\n```\n1. ASSESS   what can actually run in parallel\n2. CREATE   TeamCreate(team_name: \"descriptive-name\")\n3. SPAWN    teammates with self-contained prompts (see spawn-templates.md)\n4. TASK     create tasks with clear ownership + acceptance criteria\n5. ASSIGN   tasks to teammates\n6. WAIT     stop making tool calls — enter the WAIT state (below)\n7. PROCESS  handle messages as they arrive — checkpoint, reassign, unblock\n8. COMPLETE shut down teammates, TeamDelete\n```\n\nA framework, not a rigid script — step 7 may loop many times or once. Adapt.\n\n---\n\n## The WAIT Pattern (non-negotiable)\n\nThe single most important rule. Get it wrong and the team is useless.\n\n**Messages arrive only when the lead is idle** — not making tool calls. During an active turn they queue indefinitely.\n\n```\nAfter spawning + assigning:\n1. Finish ALL tool calls\n2. Emit a brief text summary and STOP\n3. Teammate messages arrive as new turns\n4. Process one — then STOP again for the next\n```\n\nThe rhythm is **act → stop → receive → act → stop**. Make tool calls while \"waiting\" and you will never receive teammate messages.\n\n---\n\n## File Ownership (when teammates write code)\n\n**No two teammates may edit the same file** — there is no merge resolution between concurrent teammates. Before spawning, write an ownership manifest: each teammate gets a disjoint set of paths. If a task crosses zones, split it into two tasks with a dependency.\n\nFor read-only work (audit, research), ownership doesn't apply — many agents can read the same files safely.\n\n---\n\n## Model Tiers\n\nMatch the model to the cognitive load:\n\n| Tier | Use for |\n|-|-|\n| **opus** | Writing code, reviewing\u002Fauditing, planning — anything needing judgment or correctness reasoning |\n| **sonnet** | Exploration and context-gathering — near-opus accuracy at lower latency |\n| **haiku** | Mechanical coverage only — rote navigation, trivial lookups where being wrong is cheap |\n\nCatching one subtle bug pays for an opus reviewer many times over; don't cut the model on judgment work.\n\n---\n\n## Designing the Team\n\n- **Teammates have everything you have** — CLAUDE.md, skills, tools, git. Don't paste skill content into prompts; tell them which skill to invoke. Don't extract patterns; they can explore themselves.\n- **Name by role, not number** — `frontend-coder`, `schema-auditor`, `scenario-runner` — so tasks and messages stay legible.\n- **Task descriptions are the main lever** — specific files, acceptance criteria, and ownership boundaries produce good work; vague tasks produce vague work.\n- **Absolute paths everywhere** — a teammate's cwd is not the repo root.\n- Start smaller than you think: 3-5 focused teammates beat five scattered ones.\n\nSee `spawn-templates.md` for ready-made role skeletons — use, modify, or replace them.\n\n### Data-dependent teams\n\nWhen teammates share data (E2E, integration), the lead prepares it **before** spawning:\n1. Ensure the shared data\u002Ffixtures exist — don't assume.\n2. Smoke-test ONE unit before deploying the full team; fix a BLOCKED result before wasting more agents on it.\n3. One unit of work per task — agents report back, the lead decides what's next. Don't batch.\n4. Treat BLOCKED (can't even start) as a systemic signal — pause new assignments and investigate.\n\n---\n\n## Checkpoint & Recover\n\nCheckpoint after each report: `git log --oneline` for landed commits, run the appropriate validation, update `PROGRESS.md`\u002F`WORKING.md` for phase work.\n\nAfter a context reset or crash:\n1. Read the state chain (`WORKING.md`, `PROGRESS.md`).\n2. Check whether the team still exists; resume its `TaskList`, or `TeamDelete` and rebuild from remaining work.\n3. `git log` for commits since the last checkpoint — anything committed is done.\n\nCommitted work and `PROGRESS.md` are the durable checkpoints; in-flight tasks are not.\n\n---\n\n## Anti-Patterns\n\n| Don't | Why | Instead |\n|-|-|-|\n| Tool-call while waiting | Messages queue forever | STOP after assigning |\n| \"Let me just check…\" after spawning | Same blocking problem | Spawn → assign → text → STOP |\n| Two teammates on one file | No merge resolution | Split into dependent tasks |\n| Paste skill content into prompts | Teammates have skills natively | \"Invoke skill X first\" |\n| Poll TaskList in normal flow | Wasteful; messaging works when idle | TaskList only for recovery |\n| Disposable subagent for coordinated work | Subagents can't message back | Use an Agent Team |\n| Batch many units into one task | Context bloat, no incremental feedback | One unit per task |\n\n---\n\n## Related\n\n- Spawn skeletons: `spawn-templates.md`\n- Work lifecycle: the `work` skill\n- Explorer pattern & signal vocabulary: the `patterns` skill\n",{"name":5,"description":7,"user-invocable":41},true,"main","07fef474a459f2562dbae35a3baf6470c8ae530d","template\u002F.claude\u002Fskills\u002Forchestrators\u002Fteam-lead","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit","https:\u002F\u002Fgithub.com\u002FJAuto-Corp","2026-11-10 13:45:22",{"source":8,"skillName":5,"commands":49,"prompt":59,"zipAvailable":41},[50,53,56],{"tool":51,"command":52},"npx","npx skills add JAuto-Corp\u002Fjustinventit --skill team-lead",{"tool":54,"command":55},"bunx","bunx skills add JAuto-Corp\u002Fjustinventit --skill team-lead",{"tool":57,"command":58},"pnpm","pnpm dlx skills add JAuto-Corp\u002Fjustinventit --skill team-lead","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fteam-lead\u002FSKILL.md",[61,67,71,75],{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"agents\u002Fopenai.yaml","REFERENCE","参考文档",102,"https:\u002F\u002Fgithub.com\u002Fstanley50z\u002Fsz-skills\u002Fblob\u002Fmain\u002Fskills\u002Fcodebase-design\u002Fagents\u002Fopenai.yaml",{"path":68,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":69,"sourceHtmlUrl":70},"DEEPENING.md",2553,"https:\u002F\u002Fgithub.com\u002Fstanley50z\u002Fsz-skills\u002Fblob\u002Fmain\u002Fskills\u002Fcodebase-design\u002FDEEPENING.md",{"path":72,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":73,"sourceHtmlUrl":74},"DESIGN-IT-TWICE.md",2665,"https:\u002F\u002Fgithub.com\u002Fstanley50z\u002Fsz-skills\u002Fblob\u002Fmain\u002Fskills\u002Fcodebase-design\u002FDESIGN-IT-TWICE.md",{"path":76,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":77,"sourceHtmlUrl":78},"spawn-templates.md",4484,"https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fteam-lead\u002Fspawn-templates.md",[80,107,130,152,170,186],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":84,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":36},"01M6DHBJE2P291DTXAW6DE6ZSR","ponytail","Forces the laziest solution that actually works, simplest, shortest, most minimal. Channels a senior dev who has seen everything: question whether the task needs to exist at all (YAGNI), reach for the standard library before custom code, native platform features before dependencies, one line before fifty. Supports intensity levels: lite, full (default), ultra. Use on ANY coding task: writing, adding, refactoring, fixing, reviewing, or designing code, and choosing libraries or dependencies. Also use whenever the user says \"ponytail\", \"be lazy\", \"lazy mode\", \"simplest solution\", \"minimal solution\", \"yagni\", \"do less\", or \"shortest path\", or complains about over-engineering, bloat, boilerplate, or unnecessary dependencies. Do NOT use for non-coding requests (general knowledge, prose, translation, summaries, recipes).","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Fponytail\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Fponytail\u002FSKILL.md",[],[82,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106],"line","yagni","recipes","platform","dependencies","task","do","translation","default","general","minimal","use","question","refactoring","knowledge","dev","code","libraries","writing",{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":123,"publishedAt":36},"01M6DHB61PAWVGHF53A89MJ6SM","workflow\u002Fskill","Patterns for adding or editing a skill — frontmatter contract, router + sub-file split, line budgets.","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fworkflow\u002Fskill.md",[113,118],{"categoryNo":114,"categoryName":115,"taxonomyType":116,"taxonomyTypeDesc":117},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",{"categoryNo":119,"categoryName":120,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000037","律师与法务","PROFESSION","适用职业",[124,125,126,127,128,32,129,88],"workflow skill","workflow","patterns","router","editing","contract",{"skillNo":131,"name":125,"nameZh":6,"description":132,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":133,"categories":134,"keywords":136,"publishedAt":36},"01M6DHB1GD89RW71ZYZJZJH4ZG","Meta-skill for editing the framework itself — skills, hooks, commands, rules, CLAUDE.md. Use before changing anything under .claude\u002F or the routing docs, so the change keeps the system's shape (lazy-load, forge markers, hook auto-discovery, harness).","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fworkflow\u002FSKILL.md",[135],{"categoryNo":114,"categoryName":115,"taxonomyType":116,"taxonomyTypeDesc":117},[125,137,138,139,140,141,142,99,128,143,144,145,146,147,148,149,150,151],"hooks","commands","meta-skill","shape","framework","auto","Rules","system","discovery","Meta","forge","harness","change","docs","routing",{"skillNo":153,"name":27,"nameZh":6,"description":154,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":155,"categories":156,"keywords":157,"publishedAt":36},"01M6DHAX1NCV3HXK4M1RMBWGAV","Work lifecycle orchestration: start, continue, pause, handoff, done, plus epic\u002Fsprint planning. Context gathering + goal-focused state-chain transitions.","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fwork\u002FSKILL.md",[],[27,158,159,160,161,162,163,164,165,166,167,168,33,169],"start","chain","sprint planning","transitions","done","goal","context","continue","lifecycle","state","planning","handoff",{"skillNo":171,"name":172,"nameZh":6,"description":173,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":174,"categories":175,"keywords":176,"publishedAt":36},"01M6DHASZYRB1ZZWVNQ6CN8FN3","verify","Verification and auditing. Validate acceptance scenarios, run multi-perspective audits, check type-check\u002Fbuild evidence, and gate phase\u002Fsprint completion.","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fverify\u002FSKILL.md",[],[172,177,178,179,180,181,182,183,184,185],"acceptance","run","validate","verification","type","build","gate","evidence","check",{"skillNo":187,"name":188,"nameZh":6,"description":189,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":190,"categories":191,"keywords":195,"publishedAt":36},"01M6DHAJTF40984VZYBZHFD52D","scope","Plan and design work before building. Invoke when starting an epic, scoping a sprint, or planning any Standard+ change: orient on what exists, equip the planning procedure + domain skills, then produce the SPEC\u002FSCENARIOS\u002FPROGRESS structure and converge it through an audit loop. Feeds \u002Fwork:epic-plan and the ATDD entry gate.","https:\u002F\u002Fgithub.com\u002FJAuto-Corp\u002Fjustinventit\u002Fblob\u002Fmain\u002Ftemplate\u002F.claude\u002Fskills\u002Forchestrators\u002Fscope\u002FSKILL.md",[192],{"categoryNo":193,"categoryName":194,"taxonomyType":116,"taxonomyTypeDesc":117},"01JBSEEDCAT000000000000015","视觉设计",[188,35,27,196,197,198,199,183,200,201,202,203,204,205,31,206,168,149,207],"progress","orient","domain","structure","scoping","before building","plan","loop","equip","feeds","produce","spec"]