{"id":"dc7295d1-8806-4f3f-a951-9ec94afc92a5","entityType":"agent","slug":"clawhub-beyound87-team-builder","name":"Team Builder","canonicalUrl":"https://www.xpersona.co/agent/clawhub-beyound87-team-builder","canonicalPath":"/agent/clawhub-beyound87-team-builder","generatedAt":"2026-10-10T05:17:51.521Z","source":"CLAWHUB","claimStatus":"UNCLAIMED","verificationTier":"NONE","summary":{"evidence":{"source":"editorial-content","verified":true,"confidence":"high","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":null},"description":"在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Skill: Team Builder Owner: beyound87 Summary: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Tags: latest:3.0.0 Version history: v3.0.0 | 2026-04-02T15:29:56.429Z | user 不使用acp session模式来模拟claude持久化 v2.0.6 | 2026-03-30T00:50:15.771Z | user 不使用acp session模式来模拟claude持久化 v2.0.4 | 2026-03-29T18:59:53.219Z | user 不使用a","descriptionLabel":"Technical summary","evidenceSummary":"Capability contract not published. No trust telemetry is available yet. 1.9K downloads reported by the source. Last updated 10/9/2026.","installCommand":"clawhub skill install s17dg3qthhjbatdnrgq35n8jrh83j4ba:team-builder","sourceUrl":"https://clawhub.ai/beyound87/team-builder","homepage":"https://clawhub.ai/beyound87/skills/team-builder","primaryLinks":[{"label":"View on ClawHub","url":"https://clawhub.ai/beyound87/team-builder","kind":"source"},{"label":"Homepage","url":"https://clawhub.ai/beyound87/skills/team-builder","kind":"homepage"}],"safetyScore":84,"overallRank":62,"popularityScore":58,"trustScore":null,"claimedByName":null,"isOwner":false,"seoDescription":"在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Skill: "},"coverage":{"evidence":{"source":"public-profile","verified":false,"confidence":"medium","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":null},"protocols":[{"protocol":"OPENCLEW","label":"OpenClaw","status":"self-declared","notes":"Declared in the public agent profile."}],"capabilities":[],"verifiedCount":0,"selfDeclaredCount":1,"capabilityMatrix":{"rows":[{"key":"OPENCLEW","type":"protocol","support":"unknown","confidenceSource":"profile","notes":"Listed on profile"}],"flattenedTokens":"protocol:OPENCLEW|unknown|profile"}},"adoption":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":null},"stars":null,"forks":null,"downloads":1873,"packageName":null,"latestVersion":"3.0.0","tractionLabel":"1.9K downloads"},"release":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-09T23:45:18.081Z","emptyReason":null},"lastUpdatedAt":"2026-10-09T23:45:18.082Z","lastCrawledAt":"2026-10-09T23:45:18.081Z","lastIndexedAt":null,"nextCrawlAt":"2026-10-10T23:45:18.081Z","lastVerifiedAt":null,"highlights":[{"version":"3.0.0","createdAt":"2026-04-02T15:29:56.429Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":19,"zipByteSize":63816},{"version":"2.0.6","createdAt":"2026-03-30T00:50:15.771Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":18,"zipByteSize":55995},{"version":"2.0.4","createdAt":"2026-03-29T18:59:53.219Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":18,"zipByteSize":55626},{"version":"2.0.2","createdAt":"2026-03-29T02:42:02.764Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":18,"zipByteSize":55189},{"version":"2.0.1","createdAt":"2026-03-29T02:23:00.771Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":18,"zipByteSize":55442},{"version":"2.0.0","createdAt":"2026-03-28T23:02:34.558Z","changelog":"不使用acp session模式来模拟claude持久化","fileCount":18,"zipByteSize":54384},{"version":"1.0.32","createdAt":"2026-03-28T13:13:47.752Z","changelog":"Initial release: 8-agent SaaS growth team deployer","fileCount":16,"zipByteSize":49720},{"version":"1.0.31","createdAt":"2026-03-28T06:03:51.115Z","changelog":"Initial release: 8-agent SaaS growth team deployer","fileCount":15,"zipByteSize":48301}]},"execution":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No published capability contract is available yet."},"installCommand":"clawhub skill install s17dg3qthhjbatdnrgq35n8jrh83j4ba:team-builder","setupComplexity":"low","setupSteps":["Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.","Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data."],"contract":{"contractStatus":"missing","authModes":[],"requires":[],"forbidden":[],"supportsMcp":false,"supportsA2a":false,"supportsStreaming":false,"inputSchemaRef":null,"outputSchemaRef":null,"dataRegion":null,"contractUpdatedAt":null,"sourceUpdatedAt":null,"freshnessSeconds":null},"invocationGuide":{"preferredApi":{"snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/trust"},"curlExamples":["curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/snapshot\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/contract\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/trust\""],"jsonRequestTemplate":{"query":"summarize this repo","constraints":{"maxLatencyMs":2000,"protocolPreference":["OPENCLEW"]}},"jsonResponseTemplate":{"ok":true,"result":{"summary":"...","confidence":0.9},"meta":{"source":"CLAWHUB","generatedAt":"2026-10-10T05:17:51.516Z"}},"retryPolicy":{"maxAttempts":3,"backoffMs":[500,1500,3500],"retryableConditions":["HTTP_429","HTTP_503","NETWORK_TIMEOUT"]}},"endpoints":{"dossierUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/dossier","snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-beyound87-team-builder/trust"}},"reliability":{"evidence":{"source":"runtime-metrics","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No trust, reliability, or runtime telemetry is available."},"trust":{"status":"unavailable","handshakeStatus":"UNKNOWN","verificationFreshnessHours":null,"reputationScore":null,"p95LatencyMs":null,"successRate30d":null,"fallbackRate":null,"attempts30d":null,"trustUpdatedAt":null,"trustConfidence":"unknown","sourceUpdatedAt":null,"freshnessSeconds":null},"decisionGuardrails":{"doNotUseIf":["Contract metadata is missing or unavailable for deterministic execution."],"safeUseWhen":[],"riskFlags":["missing_or_unavailable_contract","trust_data_unavailable","schema_references_missing"],"operationalConfidence":"low"},"executionMetrics":{"observedLatencyMsP50":null,"observedLatencyMsP95":null,"estimatedCostUsd":null,"uptime30d":null,"rateLimitRpm":null,"rateLimitBurst":null,"lastVerifiedAt":null,"verificationSource":null},"runtimeMetrics":{"successRate":null,"avgLatencyMs":null,"avgCostUsd":null,"hallucinationRate":null,"retryRate":null,"disputeRate":null,"p50Latency":null,"p95Latency":null,"lastUpdated":null}},"benchmarks":{"evidence":{"source":"no-benchmark-data","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No benchmark suites or observed failure patterns are available."},"suites":[],"failurePatterns":[]},"artifacts":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"high","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":null},"readme":"Skill: Team Builder\n\nOwner: beyound87\n\nSummary: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。\n\nTags: latest:3.0.0\n\nVersion history:\n\nv3.0.0 | 2026-04-02T15:29:56.429Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv2.0.6 | 2026-03-30T00:50:15.771Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv2.0.4 | 2026-03-29T18:59:53.219Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv2.0.2 | 2026-03-29T02:42:02.764Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv2.0.1 | 2026-03-29T02:23:00.771Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv2.0.0 | 2026-03-28T23:02:34.558Z | user\n\n不使用acp session模式来模拟claude持久化\n\nv1.0.32 | 2026-03-28T13:13:47.752Z | user\n\nInitial release: 8-agent SaaS growth team deployer\n\nv1.0.31 | 2026-03-28T06:03:51.115Z | user\n\nInitial release: 8-agent SaaS growth team deployer\n\nv1.0.30 | 2026-03-28T05:33:57.511Z | user\n\nInitial release: 8-agent SaaS growth team deployer\n\nv1.0.24 | 2026-03-14T08:20:34.916Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.22 | 2026-03-11T16:45:16.511Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.21 | 2026-03-11T16:34:16.587Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.20 | 2026-03-11T16:27:12.285Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.19 | 2026-03-11T16:14:13.628Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.18 | 2026-03-09T16:49:35.802Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.17 | 2026-03-07T12:22:34.573Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.16 | 2026-03-07T00:44:13.401Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.15 | 2026-03-02T15:51:12.987Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.14 | 2026-03-01T22:56:05.752Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.13 | 2026-03-01T22:44:47.979Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.12 | 2026-03-01T13:15:15.269Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.11 | 2026-03-01T12:41:18.660Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.10 | 2026-03-01T12:35:14.616Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.9 | 2026-03-01T10:37:01.748Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.8 | 2026-03-01T07:54:38.319Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.7 | 2026-03-01T07:39:29.727Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.6 | 2026-03-01T07:33:14.256Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.5 | 2026-03-01T07:27:56.861Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.4 | 2026-03-01T07:22:11.315Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.3 | 2026-03-01T07:08:44.998Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.2 | 2026-03-01T06:57:22.277Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.1 | 2026-03-01T03:56:12.595Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nv1.0.0 | 2026-03-01T03:52:26.732Z | user\n\nInitial release: 7-agent SaaS growth team deployer\n\nArchive index:\n\nArchive v3.0.0: 19 files, 63816 bytes\n\nFiles: README.md (7357b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (17777b), references/soul-templates.md (32396b), scripts/deploy.js (27535b), skill-card.md (2977b), SKILL.md (23473b), team-builder.json (394b), _meta.json (131b)\n\nFile v3.0.0:SKILL.md\n\n---\r\nname: team-builder\r\ndescription: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。\r\n---\r\n\r\n# Team Builder\r\n\r\n一条命令部署完整多 Agent 团队骨架，包含角色、信箱、看板、产品知识目录、cron 巡检、双开发轨模板。\r\n\r\n> 详细安装步骤和首次配置引导见 `README.md`（中文完整版）。\r\n\r\n## 前置条件\r\n\r\n- OpenClaw 已安装并运行\r\n- Node.js 16+\r\n- 已配置至少一个模型 provider\r\n\r\n> **此技能会修改系统配置**：`apply-config.js` 会写入 `openclaw.json`，`create-crons.*` 会创建 cron 任务。均需手动执行，不自动运行。\r\n\r\n### Internal Dispatch Protocol (MANDATORY)\r\n- **Standard agents** (product-lead, growth-lead, intel-analyst, devops, etc.): `sessions_spawn(runtime=\"subagent\", mode=\"run\")` — **禁止带 `streamTo`**\r\n- **ACP agents** (fullstack-dev): `sessions_spawn(runtime=\"acp\")` — 可带 `streamTo=\"parent\"`\r\n- 结果通过 auto-announce 推送；不要轮询 `sessions_list` 或 `subagents list`\r\n- chief-of-staff 是群内唯一入口；其他 agent 内部 spawn，不直接监听群聊\r\n- 详见 `shared/knowledge/team-workflow.md` 零章\r\n\r\n### 参谋长执行边界（MANDATORY）\r\n- **参谋长不下地干活**：任何多步骤任务（编码、调研、分析、内容、部署）→ 全部派给对应 agent\r\n- **参谋长不亲自开子代理干活**：spawn 子代理执行具体业务 = 等价于自己干活\r\n- **参谋长做且只做**：任务拆解（L1-L4 复杂度判断）、编排、派活、监控、结果汇总\r\n- 先执行后汇报；默认输出只保留：已做什么 / 拿到什么结果 / 卡在哪里\r\n\r\n### 子代理使用规则（MANDATORY，全团队适用）\r\n- **优先主 agent 自己干**：干得过来时不开子代理\r\n- **分析透再派**：开子代理前必须先分析清楚边界/依赖/输入输出\r\n- **子代理只做原子任务**：一句话说清、执行完就结束，不做判断或策略决策\r\n- **主 agent 全权负责**：判断、策略、经验积累——子代理不做这三件\r\n- **子代理结果由主 agent 汇总**后再上报参谋长\r\n\r\n### Credentials involved\r\n- **Telegram bot tokens** (optional) -- stored in openclaw.json, used for agent-to-Telegram binding\r\n- **Model API keys** -- must already be configured in your OpenClaw model providers (not handled by this skill)\r\n\r\n### Recommended\r\n- Review generated `apply-config.js` before running\r\n- Check the backup of openclaw.json after running\r\n- Test with 2-3 agents before enabling all cron jobs\r\n\r\n## Team Architecture\r\n\r\nDefault reference architecture for a SaaS/growth multi-agent team (customizable to 2-10 agents):\r\n\r\n```\r\nCEO\r\n |-- Chief of Staff (dispatch + strategy + efficiency)\r\n |-- Data Analyst (data + user research)\r\n |-- Growth Lead (GEO + SEO + community + social media)\r\n |-- Content Chief (strategy + writing + copywriting + i18n)\r\n |-- Intel Analyst (competitor monitoring + market trends)\r\n |-- Product Lead (product management + tech architecture)\r\n |-- DevOps (delivery / deploy / environment / acceptance)\r\n |-- Fullstack Dev (implementation / module deep dive / ACP coding session)\r\n```\r\n\r\n### Multi-Team Support\r\n\r\nOne OpenClaw instance can run multiple teams:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js                  # default team\r\nnode <skill-dir>/scripts/deploy.js --team alpha      # named team \"alpha\"\r\nnode <skill-dir>/scripts/deploy.js --team beta       # named team \"beta\"\r\n```\r\n\r\nNamed teams use prefixed agent IDs (`alpha-chief-of-staff`, `beta-growth-lead`) to avoid conflicts. Each team gets its own workspace subdirectory.\r\n\r\n### Flexible Team Size\r\n\r\nThe wizard lets you select 2-10 agents from the available roles. Skip roles you don't need. The 8-agent default covers most SaaS scenarios with dual-dev routing, but you can run leaner (3-4 agents) or expand with custom roles.\r\n\r\n### Model Auto-Detection\r\n\r\nThe wizard scans your `openclaw.json` for registered model providers and auto-suggests models by role type:\r\n\r\n| Role Type | Best For | Auto-detect Pattern |\r\n|-----------|----------|-------------------|\r\n| Thinking | Strategic roles (chief, growth, content, product) | /glm-5\\|opus\\|o1\\|deepthink/i |\r\n| Execution | Operational roles (data, intel, fullstack) | /glm-4\\|sonnet\\|gpt-4/i |\r\n| Fast | Lightweight tasks | /flash\\|haiku\\|mini/i |\r\n\r\nYou can always override with manual model IDs.\r\n\r\n## Setup / Config / Scripts\r\n\r\n### Required Inputs\r\n- Team name\r\n- Workspace dir\r\n- Timezone\r\n- Morning brief hour\r\n- Evening brief hour\r\n- Thinking model\r\n- Execution model\r\n- CEO title\r\n\r\n### Optional Inputs\r\n- Telegram user ID\r\n- Telegram bot tokens\r\n- Proxy\r\n- ACP coding agent(给 fullstack-dev 使用)\r\n\r\n### Core Scripts\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js\r\nnode <workspace-dir>/apply-config.js\r\npowershell <workspace-dir>/create-crons.ps1\r\nbash <workspace-dir>/create-crons.sh\r\nopenclaw gateway restart\r\n```\r\n\r\n### Execution Priority\r\n- First: matched execution skill (for coding work, `coding-lead` if loaded)\r\n- Second: agent-role fallback when no matching skill is loaded\r\n- Third: templates/README explain boundaries and ownership only; they should not override matched skills\r\n\r\n### Context File Hygiene\r\n- Active context files live under `<project>/.openclaw/`\r\n- Reuse one context file per active code chain when possible\r\n- Naming pattern: `context-<task-slug>.md`\r\n- Active context file cap per project: **60**\r\n- Context-file lifecycle window per project: **100 total files** across active + archive\r\n- Completed or stale files should be deleted or moved to `.openclaw/archive/`\r\n\r\n### Current Dual-Dev Standard\r\n- fullstack-dev:实现、模块深挖、开发文档、接口文档、Claude coding 执行;默认 coding skill 可采用 `coding-lead`,其中 simple 任务直做,medium 倾向 Claude ACP `run` 或 direct acpx,complex 通过现有会话连续协作 + 上下文文件推进,不把 ACP `session` 持久线程作为正式主路径;context 活跃上限 60、生命周期总窗口 100;并行允许但必须先定义边界,总上限 5 个工作单元\r\n- devops:交付、部署、环境、回归、冒烟、自动QA、发布门禁\r\n- product-lead:澄清、PRD、验收标准,不完整不得派工\r\n- chief-of-staff:路由、裁决、控制 token 浪费\r\n\r\n## 部署流程\r\n\r\n> 完整步骤见 `README.md`。以下是关键参数选取对照表。\r\n\r\n### 必填参数\r\n\r\n| 参数 | 默认值 | 说明 |\r\n|--------|--------|------|\r\n| teamName | Alpha Team | 团队名 |\r\n| workspaceDir | `~/.openclaw/workspace-team` | 工作区路径 |\r\n| timezone | Asia/Shanghai | cron 时区 |\r\n| morningHour | 8 | 晨报时间 |\r\n| eveningHour | 18 | 晚报时间 |\r\n| thinkingModel | 自动检测 | 策略型角色（chief/product/growth/content）|\r\n| executionModel | 自动检测 | 执行型角色（devops/fullstack/intel/data）|\r\n| ceoTitle | Boss | CEO 称呼 |\r\n\r\n### 可选参数\r\n- `roles`：自定义角色列表（默认全郥 8 个）\r\n- `roleNames`：自定义角色名称（如中文起名）\r\n- `--team <prefix>`：多团队并存时用于隔离角色 ID\r\n- Telegram bot tokens（可选，配置后自动写入 openclaw.json）\r\n\r\n### 核心命令\r\n```bash\r\n# 交互式生成\r\nnode <skill-dir>/scripts/deploy.js\r\n\r\n# 非交互生成\r\nnode <skill-dir>/scripts/deploy.js --config team-builder.json\r\n\r\n# 验证生成结果\r\nnode <skill-dir>/scripts/deploy.js --verify --config team-builder.json\r\n\r\n# 应用配置（写入 openclaw.json）\r\nnode <workspace-dir>/apply-config.js\r\n\r\n# 创建 cron\r\npowershell <workspace-dir>/create-crons.ps1  # Windows\r\nbash <workspace-dir>/create-crons.sh          # macOS/Linux\r\n\r\n# 重启\r\nopenclaw gateway restart\r\n```\r\n\r\n`--verify` 检查生成物是否包含双开发模型、角色归属、cron 条目。\r\n\r\n完整安装步骤见 `README.md`。\r\n\r\n## Cron Schedule\r\n\r\n| Offset | Agent | Task | Frequency |\r\n|--------|-------|------|-----------|\r\n| H-1 | Data Analyst | Data + user feedback | Daily |\r\n| H-1 | Intel Analyst | Competitor scan | Mon/Wed/Fri |\r\n| H | Chief of Staff | Morning brief (announced) | Daily |\r\n| H+1 | Growth Lead | GEO + SEO + community | Daily |\r\n| H+1 | Content Chief | Weekly content plan | Monday |\r\n| H+2 | DevOps | Delivery / environment / Deep Dive / acceptance | Daily |\r\n| H+10 | Chief of Staff | Evening brief (announced) | Daily |\r\n\r\n(H = morning brief hour)\r\n\r\n## Generated File Structure\r\n\r\n```\r\n<workspace>/\r\n├── AGENTS.md, SOUL.md, USER.md  (auto-injected)\r\n├── apply-config.js, create-crons.ps1/.sh, README.md\r\n├── agents/<8 agent dirs>/       (SOUL.md + MEMORY.md + memory/)\r\n└── shared/\r\n    ├── briefings/, decisions/, inbox/ (v2: with status tracking)\r\n    ├── status/team-dashboard.md     (chief-of-staff maintains, all agents read first)\r\n    ├── data/                        (public data pool, data-analyst writes, all read)\r\n    ├── kanban/, knowledge/\r\n    └── products/\r\n        ├── _index.md                (product matrix overview)\r\n        ├── _template/               (knowledge directory template)\r\n        └── {product}/               (per-product knowledge, up to 20 files)\r\n            ├── overview.md, architecture.md, database.md, api.md, routes.md\r\n            ├── models.md, services.md, frontend.md, auth.md, integrations.md\r\n            ├── jobs-events.md, config-env.md, dependencies.md, devops.md\r\n            ├── test-coverage.md, tech-debt.md, domain-flows.md, data-flow.md\r\n            ├── i18n.md, changelog.md, notes.md\r\n```\r\n\r\n\r\n\r\n## Knowledge Governance\r\n\r\nEach shared knowledge file has a designated owner. Only the owner agent updates it; others read only.\r\n\r\n| File | Owner | Update Trigger |\r\n|------|-------|---------------|\r\n| geo-playbook.md | growth-lead | After GEO experiments/discoveries |\r\n| seo-playbook.md | growth-lead | After SEO experiments |\r\n| competitor-map.md | intel-analyst | After each competitor scan |\r\n| content-guidelines.md | content-chief | After proven writing patterns |\r\n| user-personas.md | data-analyst | After new user insights |\r\n| tech-standards.md | product-lead | After architecture decisions |\r\n\r\n### Update Protocol\r\nWhen updating a knowledge file, the owner must:\r\n1. Add a dated entry at the top: `## [YYYY-MM-DD] <what changed>`\r\n2. Include the reason and data evidence\r\n3. Never delete existing entries without CEO approval (append, don't replace)\r\n\r\n### Chief of Staff Governance\r\nThe chief-of-staff monitors knowledge file health during weekly reviews:\r\n- Are files being updated regularly?\r\n- Any conflicting information between files?\r\n- Any stale entries that should be archived?\r\n\r\n## Self-Evolution Pattern\r\n\r\nAgents improve their own strategies over time through a feedback loop:\r\n\r\n```\r\n1. Execute task (cron or inbox triggered)\r\n2. Collect results (data, metrics, outcomes)\r\n3. Analyze: what worked vs what didn't\r\n4. Update knowledge files with proven strategies (with evidence)\r\n5. Next execution reads updated knowledge → better performance\r\n```\r\n\r\nThis is NOT the agent randomly changing rules. Updates must be:\r\n- **Data-driven**: backed by metrics or concrete outcomes\r\n- **Incremental**: append new findings, don't rewrite everything\r\n- **Traceable**: dated with evidence so others can verify\r\n\r\n### What Agents Can Self-Update\r\n- Their own knowledge files (per ownership table above)\r\n- Their own MEMORY.md (lessons learned, decisions)\r\n- shared/data/ outputs (data-analyst only)\r\n\r\n### What Requires CEO Approval\r\n- shared/decisions/active.md (strategy changes)\r\n- Adding/removing agents or changing team architecture\r\n- External publishing or spending decisions\r\n\r\n## Public Data Layer\r\n\r\nThe `shared/data/` directory serves as a read-only data pool for all agents:\r\n\r\n- **data-analyst** writes: daily metrics, user feedback summaries, anomaly alerts\r\n- **All agents** read: to inform their own decisions\r\n- Format: structured markdown or JSON, dated filenames (e.g., `metrics-2026-03-01.md`)\r\n- Retention: keep 30 days, archive older files\r\n\r\n## Project Deep Dive - Code Scanning\r\n\r\nAgents can deeply understand each SaaS product through automated code scanning. This is critical - without deep project knowledge, all team decisions are surface-level.\r\n\r\n### How It Works\r\n\r\n1. CEO adds a product to `shared/products/_index.md` (name, URL, code directory, tech stack)\r\n2. Product Lead triggers a delivery-oriented Deep Dive scan by dispatching to DevOps (via `sessions_spawn` if online, inbox as fallback)\r\n3. DevOps enters the project directory (read-only) and generates shared knowledge / delivery-oriented scan outputs\r\n4. Fullstack Dev picks up module-level deep dive or implementation follow-up when needed\r\n5. Knowledge files are generated in `shared/products/{product}/`\r\n6. All agents consume these files **via manifest-based lazy loading** (never read all at once)\r\n\r\n### Manifest-Based Lazy Loading (MANDATORY)\r\n\r\nEach product directory includes a `manifest.json` (~200 tokens) that lists all files with one-line summaries and a `taskFileMap` mapping task types to relevant files.\r\n\r\n**Agent workflow:**\r\n1. Read `_index.md` → identify which product\r\n2. Read `{product}/manifest.json` → see all files + summaries (~200 tokens)\r\n3. Based on `taskFileMap` or summaries, read only 1-3 relevant files\r\n4. Never read more than 5 product files per session\r\n\r\n**Why:** With 15+ products × 20 files each, full loading = 40K+ tokens per product. Manifest loading = 200 tokens + only what's needed.\r\n\r\n**DevOps MUST regenerate `manifest.json`** after every delivery-oriented scan (L0-L4). Fullstack Dev updates it when doing module-level follow-up that changes knowledge scope. Template in `_template/manifest.json`.\r\n\r\n### Manifest Quality Standards\r\n\r\n摘要不能为了省 token 丢掉关键信息。每条摘要须满足:\r\n- **核心文件**(database/models/services/routes/integrations):50-130字,列出关键实体名/数量/域名\r\n- **中等文件**(auth/frontend/commands/config):30-80字,点明方案和范围\r\n- **轻量文件**(changelog/notes/metrics):可以短(<20字)\r\n- **taskFileMap**:必须覆盖该产品的所有核心业务场景(不少于8个映射)\r\n- **codeStats**:必须包含文件数、行数、模型数、表数等量化指标\r\n\r\n### Product Knowledge Directory\r\n\r\nEach product gets a knowledge directory with up to 20 files + manifest:\r\n\r\n```\r\nshared/products/{product}/\r\n├── manifest.json        ← **INDEX** (~200 tokens): file list, summaries, taskFileMap\r\n├── overview.md          ← Product positioning (from _index.md)\r\n├── architecture.md      ← System architecture, tech stack, design patterns, layering\r\n├── database.md          ← Full table schema, relationships, indexes, migrations\r\n├── api.md               ← API endpoints, params, auth, versioning\r\n├── routes.md            ← Complete route table (Web + API + Console)\r\n├── models.md            ← ORM relationships, scopes, accessors, observers\r\n├── services.md          ← Business logic, state machines, workflows, validation\r\n├── frontend.md          ← Component tree, page routing, state management\r\n├── auth.md              ← Auth scheme, roles/permissions matrix, OAuth\r\n├── integrations.md      ← Third-party: payment/email/SMS/storage/CDN/analytics\r\n├── jobs-events.md       ← Queue jobs, event listeners, scheduled tasks, notifications\r\n├── config-env.md        ← Environment variables, feature flags, cache strategy\r\n├── dependencies.md      ← Key dependencies, custom packages, vulnerabilities\r\n├── devops.md            ← Deployment, CI/CD, Docker, monitoring, logging\r\n├── test-coverage.md     ← Test strategy, coverage, weak spots\r\n├── tech-debt.md         ← TODO/FIXME/HACK inventory, dead code, complexity hotspots\r\n├── domain-flows.md      ← Core user journeys, domain boundaries, module coupling\r\n├── data-flow.md         ← Data lifecycle: external → import → process → store → output\r\n├── i18n.md              ← Internationalization, language coverage\r\n├── changelog.md         ← Scan diff log (what changed between scans)\r\n└── notes.md             ← Agent discoveries, gotchas, implicit rules\r\n```\r\n\r\n### Scan Levels\r\n\r\n| Level | Scope | When | Output |\r\n|-------|-------|------|--------|\r\n| L0 Snapshot | Surface: directory tree, packages, env | First onboard | architecture, dependencies, config-env |\r\n| L1 Skeleton | Structure: DB, routes, models, components | First onboard | database, routes, api, models, frontend |\r\n| L2 Deep Dive | Logic: services, auth, jobs, integrations | On-demand per module | services, auth, jobs-events, integrations, domain-flows, data-flow |\r\n| L3 Health Check | Quality: tech debt, tests, security | Periodic / pre-release | tech-debt, test-coverage, devops |\r\n| L4 Incremental | Delta: git diff → update affected files | After code changes | changelog + targeted updates |\r\n\r\n### Content Standards\r\n\r\nKnowledge files capture not just WHAT exists but WHY:\r\n- **Design decisions**: Why this approach was chosen\r\n- **Implicit business rules**: Logic buried in code (e.g., \"orders auto-cancel after 72h\")\r\n- **Gotchas**: What breaks if you touch this module carelessly\r\n- **Cross-module coupling**: Where changing A silently breaks B\r\n- **Performance hotspots**: N+1 queries, missing indexes, bottleneck endpoints\r\n\r\n### Role Responsibilities\r\n\r\n| Role | Responsibility |\r\n|------|---------------|\r\n| Product Lead | **Clarification / PRD / acceptance**: complete clarification, PRD, user stories, acceptance criteria, and review knowledge freshness before delegating |\r\n| DevOps | **Delivery / QA gate / Deep Dive**: enter code directory for deployment-oriented scans, maintain release checklist, smoke/regression testing, auto-QA access, and generate/update shared product knowledge files |\r\n| Fullstack Dev | **Implementation / docs / Deep Dive follow-up**: continue module-level deep dive, code analysis, implementation, dev docs, interface docs, and ACP session work |\r\n| Chief of Staff | **Routing / escalation**: split implementation vs delivery tasks, prevent duplicate labor, escalate blockers |\r\n| All Agents | **Consumption**: read product knowledge before any product-related decision |\r\n\r\n### Per-Stack Auto-Detection\r\n\r\nFullstack Dev auto-detects tech stack and applies stack-specific scan strategies:\r\n- **Laravel/PHP**: migrations, route:list, Models, Services, Middleware, Policies, Jobs, Console/Kernel\r\n- **React/Vue**: components, router, stores, API client, i18n\r\n- **Python/Django/FastAPI**: models.py, urls.py, views.py, middleware, celery\r\n- **General**: tree, git log, grep TODO/FIXME, .env.example, Docker, CI, tests\r\n\r\n## Team Coordination v2\r\n\r\n### Inbox Protocol v2 (backup channel, status tracking)\r\n\r\n> **Primary dispatch**: `sessions_spawn` (real-time). Inbox is for archival, cross-session handoff, and fallback when spawn is unavailable.\r\n\r\nEvery inbox message now has a `status` field:\r\n- `pending` → `received` → `in-progress` → `done` (or `blocked`)\r\n- Chief-of-staff monitors timeouts: high>4h, normal>24h pending = intervention\r\n- Blocked >8h = escalation to CEO\r\n- Recipients MUST update status immediately upon reading\r\n\r\n### Team Dashboard (`shared/status/team-dashboard.md`)\r\n\r\nChief-of-staff maintains a \"live scoreboard\" updated every session:\r\n- 🔴 Urgent/Blocked items\r\n- 📊 Per-agent status table (last active, current task, status icon)\r\n- 📬 Unprocessed inbox summary (pending/blocked messages across all inboxes)\r\n- 🔗 Cross-agent task chain tracking (A→B→C with per-step status)\r\n- 📅 Today/Tomorrow focus\r\n\r\n**Agent 启动顺序（内置于 AGENTS.md）：**\r\n1. 确认角色\r\n2. 读 `agents/[role]/SOUL.md`\r\n3. 读 `shared/onboarding.md`（项目背景，CEO 填写）\r\n4. 读 `shared/status/team-dashboard.md`（当前状态）\r\n5. 读 `shared/decisions/active.md`（仅涉及策略时）\r\n6. 读 `shared/inbox/to-[role].md`\r\n7. 读 `agents/[role]/MEMORY.md`（仅需历史上下文时）\r\n\r\n### Chief-of-Staff as Router\r\n\r\nThe chief is upgraded from \"briefing writer\" to \"active team router\":\r\n- **Real-time dispatch**: uses `sessions_spawn(runtime=\"subagent\")` to directly wake agents and assign tasks - this is the primary dispatch method\r\n- **Blocker detection**: scans all inboxes for overdue messages\r\n- **Inbox as backup**: writes to inbox only for archival, cross-session handoff, or when agent is unreachable\r\n- **Task chain tracking**: identifies multi-agent workflows and tracks each step\r\n- **Escalation**: persistent blockers get flagged to CEO\r\n- **Runs 4x/day** (morning brief, midday patrol, afternoon patrol, evening brief)\r\n\r\n### Cron Schedule (10 jobs, up from 7)\r\n\r\n| Time | Agent | Type | Purpose |\r\n|------|-------|------|---------|\r\n| 07:00 | data-analyst | daily | Data pull + feedback scan |\r\n| 08:00 | chief-of-staff | **announce** | Morning: router scan + brief + quality |\r\n| 09:00 | growth-lead | daily | GEO/SEO/community |\r\n| 09:00 | product-lead | **daily (NEW)** | Inbox + clarification/PRD + task delegation |\r\n| 10:00 | content-chief | **daily M-F (was weekly)** | Content creation + collaboration |\r\n| 10:00 | devops | **daily (delivery track)** | Inbox + Deep Dive + delivery + QA gate |\r\n| 12:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 15:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 18:00 | chief-of-staff | **announce** | Evening: router scan + summary + next day plan |\r\n| 07:00 M/W/F | intel-analyst | 3x/week | Competitor scan |\r\n\r\n### Why These Changes Matter\r\n\r\n| Before | After | Impact |\r\n|--------|-------|--------|\r\n| Inbox = primary dispatch | Inbox = backup + spawn = primary | Real-time dispatch via spawn; inbox for archival only |\r\n| Chief 2x/day | Chief 4x/day with router role | Blockers caught within hours, not days |\r\n| Content-chief 1x/week | Daily M-F | Actually produces content |\r\n| Product-lead no cron | Daily | Knowledge governance happens |\r\n| No team dashboard | Dashboard every session | All agents know the full picture |\r\n| No timeout detection | Automatic timeout rules | Nothing falls through cracks |\r\n\r\n## Key Design Decisions\r\n\r\n- **Shared workspace** so qmd indexes everything for all agents\r\n- **Real-time spawn dispatch** as primary inter-agent communication; inbox as backup for archival and cross-session handoff\r\n- **Chief as Router** - active coordinator who dispatches via `sessions_spawn`, detects blockers, and resolves them\r\n- **Team Dashboard** - single source of truth for team-wide status, maintained by chief every session\r\n- **GEO as #1 priority** (AI search = blue ocean)\r\n- **Fullstack Dev spawns Claude Code** via ACP for complex implementation tasks\r\n- **DevOps owns delivery and QA gate** so implementation and release responsibilities stay separated\r\n- **Project Deep Dive** gives all agents deep codebase understanding, not just surface-level product overviews\r\n\r\n## Customization\r\n\r\nEdit ROLES array in `scripts/deploy.js` to add/remove agents.\r\nEdit `references/soul-templates.md` for SOUL.md templates.\r\nEdit `references/shared-templates.md` for shared file templates.\n\nFile v3.0.0:README.md\n\n# team-builder\n\n多 Agent 团队工作区生成工具。一条命令搭起完整的团队骨架，包含角色、信箱、看板、产品知识目录、cron 巡检、双开发轨模板。\n\n---\n\n## 🚀 首次使用：完整安装流程\n\n### 第一步：确认前置条件\n\n```bash\nnode -v        # 需要 Node.js 16+\nopenclaw status  # 确认 OpenClaw 已运行\n```\n\n### 第二步：运行生成脚本（交互式）\n\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js\n```\n\n脚本会依次询问：\n- 团队名称（如 `Alpha Team`）\n- 工作区路径（默认 `~/.openclaw/workspace-team`）\n- 时区（默认 `Asia/Shanghai`）\n- 晨报/晚报时间（默认 8 / 18）\n- 思考型模型（用于 chief-of-staff / product-lead / growth-lead）\n- 执行型模型（用于 devops / fullstack-dev 等）\n- CEO 称呼（如 `老板` / `Boss`）\n- 各角色名称（可直接回车用默认值）\n\n> 💡 不想交互？用 `--config`，见下方「非交互部署」。\n\n### 第三步：应用配置（写入 openclaw.json）\n\n```bash\nnode <工作区路径>/apply-config.js\n```\n\n这一步把生成的 Agent 写入 OpenClaw 配置，**必须执行**。\n\n### 第四步：创建定时任务\n\n**Windows（PowerShell）：**\n```powershell\npowershell <工作区路径>/create-crons.ps1\n```\n\n**macOS / Linux（Bash）：**\n```bash\nbash <工作区路径>/create-crons.sh\n```\n\n### 第五步：重启 Gateway\n\n```bash\nopenclaw gateway restart\n```\n\n### 第六步：CEO 填写项目上下文 ⚠️ 重要\n\n打开工作区里的 `shared/onboarding.md`，**填写以下内容**：\n\n```markdown\n## 当前项目/产品\n（是什么产品，目标用户，核心功能一句话）\n\n## 现在进展到哪步\n（已完成什么，当前在做什么）\n\n## 最近完成的里程碑\n（近期上线/交付了什么，时间）\n\n## 当前最大阻塞\n（卡在哪里，需要什么才能继续）\n\n## 技术栈/代码目录\n（主要技术栈，代码在哪个路径）\n```\n\n> ⚠️ **这一步非常重要**：所有 Agent 启动时会先读这个文件。不填的话 Agent 不知道你在做什么项目，会从零开始问你。\n> 填完后告知 chief-of-staff（直接在群里说一声），它会开始正式工作。\n\n### 第七步：验证（可选但推荐）\n\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --verify --config <配置文件路径>\n```\n\n输出结构化 JSON 检查报告，覆盖 `core / roles / fallback / workflow / hygiene / cron`。\n\n---\n\n## ⚙️ 非交互部署（推荐用于重复/CI 场景）\n\n准备配置文件 `team-builder.json`：\n\n```json\n{\n  \"teamName\": \"Alpha Team\",\n  \"workspaceDir\": \"~/.openclaw/workspace-team\",\n  \"timezone\": \"Asia/Shanghai\",\n  \"morningHour\": 8,\n  \"eveningHour\": 18,\n  \"thinkingModel\": \"claude-sonnet-4-5\",\n  \"executionModel\": \"claude-sonnet-4-5\",\n  \"ceoTitle\": \"老板\",\n  \"roles\": [\"chief-of-staff\", \"product-lead\", \"devops\", \"fullstack-dev\", \"growth-lead\", \"content-chief\", \"intel-analyst\", \"data-analyst\"],\n  \"roleNames\": {\n    \"chief-of-staff\": \"参谋长\",\n    \"product-lead\": \"产品总监\",\n    \"devops\": \"交付总监\",\n    \"fullstack-dev\": \"全栈工程师\",\n    \"growth-lead\": \"增长总监\",\n    \"content-chief\": \"内容主编\",\n    \"intel-analyst\": \"情报分析师\",\n    \"data-analyst\": \"数据分析师\"\n  }\n}\n```\n\n运行：\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --config team-builder.json\n```\n\n---\n\n## 📁 生成物说明\n\n| 文件/目录 | 说明 |\n|-----------|------|\n| `AGENTS.md` | 启动引导 + 角色对照表 |\n| `SOUL.md` | 团队价值观（可自定义） |\n| `USER.md` | CEO 信息 |\n| `shared/onboarding.md` | **CEO 填写项目背景**（新增，首次必填） |\n| `shared/status/team-dashboard.md` | 团队状态面板（chief-of-staff 维护） |\n| `shared/decisions/active.md` | CEO 决策与指令 |\n| `shared/products/_index.md` | 产品矩阵目录 |\n| `shared/inbox/to-[role].md` | 各角色信箱 |\n| `shared/knowledge/team-workflow.md` | 团队工作流规范 |\n| `agents/[role]/SOUL.md` | 各角色身份与职责 |\n| `agents/[role]/MEMORY.md` | 各角色长期记忆 |\n| `apply-config.js` | 写入 openclaw.json（手动执行） |\n| `create-crons.ps1 / .sh` | 创建 cron 定时任务（手动执行） |\n| `references/coding-behavior-fallback.md` | fullstack-dev 编码行为兜底 |\n\n---\n\n## 🧠 Agent 启动顺序（已内置于 AGENTS.md）\n\n每个 Agent 启动时自动按此顺序读文件：\n\n1. 确认自己的角色身份\n2. 读 `agents/[role]/SOUL.md`\n3. **读 `shared/onboarding.md`**（项目背景，CEO 填写）\n4. **读 `shared/status/team-dashboard.md`**（当前状态）\n5. 读 `shared/decisions/active.md`（仅当涉及策略/决策时）\n6. 读 `shared/inbox/to-[role].md`（自己的收件箱）\n7. 读 `agents/[role]/MEMORY.md`（仅当需要历史上下文时）\n\n---\n\n## 🔄 双开发轨模型\n\n```\nproduct-lead  ←→  需求澄清 / PRD / 验收标准 / 路由\n     ↓\ndevops           ←→  部署 / 环境 / QA gate / 发布\nfullstack-dev    ←→  项目内实现 / 模块深挖 / 连续编码\n```\n\n编码执行分层：\n- 简单任务 → 直接执行\n- 中等任务 → ACP run / direct acpx\n- 复杂任务 → 现有 ACP session 连续性 + context file\n\n---\n\n## 📋 可选角色说明\n\n| 角色 ID | 职责 | 模型类型 |\n|---------|------|---------|\n| `chief-of-staff` | 路由器 + 调度 + 策略（**必选**） | 思考型 |\n| `product-lead` | 产品管理 + 需求澄清 + 验收 | 思考型 |\n| `devops` | 交付 + 部署 + QA | 执行型 |\n| `fullstack-dev` | 代码实现 + 模块深挖 | 执行型 |\n| `growth-lead` | GEO + SEO + 社区 + 社媒 | 思考型 |\n| `content-chief` | 内容策略 + 写作 + 本地化 | 思考型 |\n| `intel-analyst` | 竞品情报 + 市场趋势 | 执行型 |\n| `data-analyst` | 数据分析 + 用户研究 | 执行型 |\n\n---\n\n## 🔧 常用命令速查\n\n```bash\n# 交互式生成\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js\n\n# 非交互生成\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --config team-builder.json\n\n# 验证生成结果\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --verify --config team-builder.json\n\n# 应用配置\nnode <工作区>/apply-config.js\n\n# 重启 gateway\nopenclaw gateway restart\n```\n\n---\n\n## ✅ 安装后检查清单\n\n- [ ] `apply-config.js` 已执行，openclaw.json 已更新\n- [ ] `create-crons.ps1 / .sh` 已执行，定时任务已创建\n- [ ] `openclaw gateway restart` 已执行\n- [ ] `shared/onboarding.md` 已填写项目背景\n- [ ] `shared/products/_index.md` 已填写产品信息\n- [ ] `shared/decisions/active.md` 已填写 CEO 指令\n- [ ] 已在群里告知 chief-of-staff 开始工作\n\n---\n\n## 相关文件\n\n- `SKILL.md` — Agent 技能描述\n- `scripts/deploy.js` — 核心生成脚本\n- `references/shared-templates.md` — 共享文件模板\n- `references/soul-templates.md` — 角色 SOUL 模板\n- `references/coding-behavior-fallback.md` — 编码行为兜底规则\n- `references/agent-refs/` — 各角色方法论参考文件\n\n---\n\n## 最近更新\n\n- **2026-04-02**：新增 `shared/onboarding.md` 项目背景引导文件；`AGENTS.md` 启动顺序加入读 onboarding 步骤；README 重写为中文完整安装指引\n- **2026-03-29**：`deploy.js --verify` 扩展为结构化 JSON 检查报告；生成物自动带 `team-workflow.md`\n\nFile v3.0.0:_meta.json\n\n{\n  \"ownerId\": \"kn761rne5jajzfg0d8h6nge4hh8220cq\",\n  \"slug\": \"team-builder\",\n  \"version\": \"3.0.0\",\n  \"publishedAt\": 1775143796429\n}\n\nFile v3.0.0:references/agent-refs/chief-of-staff/dashboard-template.md\n\n# 团队面板模板\n\n```markdown\n# 团队状态面板\n> 参谋长维护 | 最后更新: YYYY-MM-DD HH:MM\n\n## 🔴 紧急/阻塞\n- [描述] — 涉及: [agent] — 原因: [原因] — 已持续: [时长]\n\n## 📊 各agent状态\n| Agent | 上次活跃 | 当前任务 | 状态 |\n|-------|---------|---------|------|\n| 参谋长 | 刚才 | 面板+简报 | ✅ |\n| 产品总监 | | | |\n| 增长总监 | | | |\n| 内容主编 | | | |\n| 数据分析 | | | |\n| 情报分析 | | | |\n| 全栈工程 | | | |\n\n状态: ✅正常 🔄进行中 🔴阻塞 ⏳等待 💤失联(>48h)\n\n## 📬 未处理信箱\n| 消息 | 发→收 | 优先级 | 已等待 | 状态 |\n\n## 🔗 跨agent任务链\n- [任务]: A(✅) → B(🔄) → C(⏳)\n\n## 📅 今日/明日重点\n1. ...\n```\n\nFile v3.0.0:references/agent-refs/chief-of-staff/strategy-methodology.md\n\n# 矩阵策略方法论\n\n## 产品组合评估框架\n\n### 产品健康度评分卡（每产品，月度更新）\n\n| 维度 | 指标 | 权重 | 评分(1-5) | 说明 |\n|------|------|:----:|:---------:|------|\n| **流量** | 月UV增长率 | 20% | | >20%=5, 10-20%=4, 0-10%=3, 持平=2, 下降=1 |\n| **留存** | D7回访率 | 20% | | >40%=5, 30-40%=4, 20-30%=3, 10-20%=2, <10%=1 |\n| **变现** | 收入/利润趋势 | 15% | | 盈利=5, 接近盈亏=4, 有收入=3, 无收入=2, 亏损=1 |\n| **竞争** | 差异化强度 | 15% | | 独占优势=5, 明显差异=4, 有差异=3, 同质=2, 落后=1 |\n| **效率** | 投入产出比 | 15% | | 高ROI=5, 正向=4, 中等=3, 低=2, 负=1 |\n| **战略** | 对矩阵协同价值 | 15% | | 核心枢纽=5, 高协同=4, 有协同=3, 独立=2, 拖累=1 |\n\n**总分 = Σ(评分 × 权重)**\n\n### 产品分类（简化BCG矩阵）\n\n| 类型 | 条件 | 策略 |\n|------|------|------|\n| ⭐ **明星** | 高增长 + 高份额（或潜力） | 全力投入，抢市场 |\n| 🐄 **现金牛** | 低增长 + 高份额 | 维护为主，收割利润，少投入 |\n| ❓ **问号** | 高增长 + 低份额 | 限期验证：3个月达标→升星，否则降级 |\n| 🐕 **瘦狗** | 低增长 + 低份额 | 最低维护或考虑下线，资源释放 |\n\n### 产品组合仪表盘模板\n```markdown\n# 产品矩阵健康度 — {YYYY}-{MM}\n\n| 产品 | 类型 | 健康度 | 本月重点 | 资源分配 |\n|------|------|:------:|---------|:--------:|\n| Richlyva | ❓问号 | 3.2/5 | 冷启动验证 | 40% |\n| ... | | | | |\n\n## 矩阵级风险\n- {跨产品风险}\n\n## 协同机会\n- {跨产品引流/共享组件}\n```\n\n## 资源分配模型\n\n### 核心原则\nCEO一个人 + AI团队，资源 = **CEO时间 + Agent算力 + 注意力**。\n\n### 分配公式\n```\n产品资源占比 = (健康度得分 × 战略权重) / Σ(所有产品的 健康度×权重)\n```\n\n其中**战略权重**由CEO指定（1-3倍）：\n- 3× = CEO明确标注的战略重点\n- 2× = 有明确增长潜力\n- 1× = 维护型\n\n### 分配约束\n| 规则 | 约束 |\n|------|------|\n| 单产品上限 | ≤60%（避免all-in风险） |\n| 单产品下限 | ≥5%（维护最低限度） |\n| 新产品试验 | 独立预留10-15%（不从现有产品扣） |\n| 调整频率 | 月度review，CEO批准后生效 |\n\n### Agent工时分配（参考）\n| 角色 | 分配逻辑 |\n|------|---------|\n| fullstack-dev | 按产品资源占比分配coding时间 |\n| content-chief | 按产品资源占比分配内容产出 |\n| growth-lead | 聚焦当前AARRR最弱环节的产品 |\n| data-analyst | 各产品均匀覆盖（数据不能偏科） |\n| intel-analyst | 聚焦⭐明星和❓问号的竞品 |\n\n### 资源重分配触发条件\n| 信号 | 动作 |\n|------|------|\n| 产品健康度连续2月下降 | 减少投入，除非CEO坚持 |\n| 产品健康度突破4.0 | 考虑加大投入（明星候选） |\n| ❓问号产品3月未达标 | 降为🐕瘦狗，释放资源 |\n| CEO新增产品 | 从预留池分配，不挤占现有 |\n| 竞品重大威胁 | 临时增加相关产品投入（参谋长判断+CEO批准） |\n\n## 跨产品协同策略\n\n### 协同类型\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **流量协同** | A产品用户引导到B产品 | 影视站→工具站推荐 |\n| **技术协同** | 共享组件/服务 | 统一认证、共享UI库 |\n| **数据协同** | 共享用户画像/行为数据 | 用户兴趣标签跨产品复用 |\n| **品牌协同** | 统一品牌矩阵 | 主品牌→子品牌信任传递 |\n| **SEO协同** | 站群互链+内容交叉引用 | 但要注意搜索引擎惩罚风险 |\n\n### 协同评估\n每次发现协同机会时评估：\n- **收益**: 预期带来多少增量（流量/用户/收入）\n- **成本**: 需要多少开发/运营投入\n- **风险**: 是否有负面影响（如SEO惩罚、品牌稀释）\n- **依赖**: 需要几个Agent配合，协调成本高不高\n\n## 定价策略分析框架\n\n### 定价三角\n```\n        价值感知\n       /        \\\n      /          \\\n竞品价格 ——— 成本底线\n```\n\n### 定价方法选择\n| 阶段 | 方法 | 说明 |\n|------|------|------|\n| 冷启动 | 免费/Freemium | 先要用户，不要收入 |\n| 有留存 | 功能分级 | 基础免费 + 高级付费 |\n| 有品牌 | 价值定价 | 按用户获得的价值定价 |\n| 成熟期 | 竞品对标 | 参考竞品定价+差异化溢价 |\n\n### 价格测试框架\n- A/B测试不同价格点（需要足够流量）\n- 用户调研：\"你愿意为X付多少\"\n- 竞品定价对标分析（intel-analyst 提供数据）\n- Van Westendorp价格敏感度（如果能做调研的话）\n\nFile v3.0.0:references/agent-refs/content-chief/methodology.md\n\n# 内容方法论\n\n## 内容漏斗匹配（Content-Funnel Fit）\n\n每篇内容明确服务哪个用户旅程阶段：\n\n| 阶段 | 用户意图 | 内容类型 | 示例 |\n|------|---------|---------|------|\n| **TOFU** 认知 | \"有这类工具吗？\" | 科普、趋势、问题解答 | \"2026年影视数据库对比\" |\n| **MOFU** 考虑 | \"哪个适合我？\" | 对比评测、教程、案例 | \"Richlyva vs IMDb 角色图谱对比\" |\n| **BOFU** 决策 | \"我要用这个\" | 上手指南、API文档 | \"5分钟搭建影视宇宙\" |\n\n**比例约 5:3:2（TOFU:MOFU:BOFU）**\n\n## 内容日历与排期\n\n### 发布节奏（冷启动期）\n| 类型 | 频率 | 负责 | 说明 |\n|------|------|------|------|\n| 长文（>1500词） | 2篇/周 | 内容主编主导 | SEO/GEO核心资产 |\n| 短文/列表（500-1000词） | 2篇/周 | 可委托或模板化 | 长尾关键词覆盖 |\n| 社区帖（Reddit/HN/Quora） | 3条/周 | 配合growth-lead | 引流+品牌曝光 |\n| FAQ/更新日志 | 按需 | 配合product-lead | 站内SEO+用户信任 |\n\n### 选题流程（4步）\n1. **关键词池** → 从 SEO/GEO 关键词研究中选取本周目标词\n2. **意图匹配** → 判断搜索意图（信息型/导航型/交易型）→ 匹配漏斗阶段\n3. **竞品内容差距** → intel-analyst 竞品数据中找对手没覆盖的角度\n4. **排期** → 写入 `shared/knowledge/content-plan-{YYYY}-W{nn}.md`\n\n### 内容日历模板\n```markdown\n## 内容计划 {YYYY}-W{nn}\n| 日期 | 标题 | 漏斗 | 目标关键词 | 渠道 | 状态 |\n|------|------|------|-----------|------|------|\n| 周一 | ... | TOFU | ... | Blog | 草稿/发布/推迟 |\n```\n\n## SEO/GEO 关键词研究流程\n\n### 第1步：种子词收集\n- 产品核心功能词（character map, movie universe, film database）\n- 用户痛点词（从 data-analyst 反馈/搜索query）\n- 竞品品牌词+替代词（IMDb alternative, letterboxd vs）\n- AI搜索常见问法（best tool for..., how to find...）\n\n### 第2步：扩展与验证\n| 工具/方法 | 用途 | 成本 |\n|-----------|------|------|\n| Google Suggest / People Also Ask | 长尾扩展 | 免费 |\n| Bing Webmaster | 实际搜索数据 | 免费 |\n| 竞品页面H1/H2/Title扫描 | 发现覆盖差距 | 免费 |\n| Reddit/Quora热门问题 | 真实用户措辞 | 免费 |\n| Google Trends | 趋势验证 | 免费 |\n\n### 第3步：关键词评分\n| 因子 | 说明 | 权重 |\n|------|------|:----:|\n| 搜索量预估 | 高/中/低（无精确数据时用趋势判断） | 30% |\n| 竞争难度 | SERP前10是大站还是小站 | 25% |\n| 意图匹配 | 和产品功能的相关度 | 25% |\n| GEO潜力 | AI搜索是否会引用此类内容 | 20% |\n\n### 第4步：关键词档案\n```markdown\n## 关键词: {keyword}\n- 搜索量预估: 高/中/低\n- 竞争难度: 高/中/低\n- 意图: 信息型/导航型/交易型\n- 漏斗阶段: TOFU/MOFU/BOFU\n- SERP现状: {前3结果概述}\n- GEO机会: {AI搜索是否已有引用}\n- 优先级: P1/P2/P3\n- 计划内容: {标题草案}\n```\n\n## 内容效果评估框架\n\n### 发布后追踪（分阶段）\n| 时间点 | 检查项 | 合格线 |\n|--------|--------|--------|\n| 发布即时 | Google索引（site:命令） | 48h内收录 |\n| 7天 | Search Console曝光量 | >0（有曝光） |\n| 14天 | 排名位置 | 目标词进前50 |\n| 30天 | 点击量、平均排名 | 目标词进前20 |\n| 60天 | 自然流量占比 | 该页有稳定流量 |\n\n### 内容健康度分级\n| 等级 | 标准 | 行动 |\n|------|------|------|\n| 🟢 健康 | 排名稳定/上升，有流量 | 保持，考虑扩展内链 |\n| 🟡 待优化 | 排名11-30，有曝光无点击 | 优化标题/摘要/结构 |\n| 🔴 失败 | 60天无排名无流量 | 分析原因：选题错？内容薄？竞争强？ |\n| ⚫ 过时 | >6个月未更新且流量下降 | 更新或合并 |\n\n### 内容ROI评估\n```\n内容价值 = (自然流量 × 预估CPC) / 创作时间(小时)\n```\n无精确数据时用定性判断：这篇内容带来了注册/咨询/品牌曝光吗？\n\n## 内容复用策略（1→N）\n\n一篇核心长文可拆解为：\n\n| 衍生形式 | 渠道 | 改造量 |\n|---------|------|--------|\n| 精华摘要（300字） | Reddit/HN帖子 | 低 |\n| FAQ拆条 | Quora回答 | 低 |\n| 数据可视化 | Twitter/社媒图 | 中 |\n| 对比表格 | 独立落地页 | 中 |\n| 视频脚本大纲 | YouTube（未来） | 高 |\n| Newsletter片段 | 邮件（未来） | 低 |\n\n**规则：核心长文先发博客 → 等收录后 → 再拆分发各渠道（避免重复内容惩罚）**\n\n## 写作检查清单（发布前必检）\n- [ ] 标题含关键词且吸引点击\n- [ ] 开头30字回答搜索意图（AI搜索截取用）\n- [ ] H2/H3 含关键词变体\n- [ ] FAQ 区块 3-5 个（Schema+GEO友好）\n- [ ] 内链 ≥2 个\n- [ ] CTA 明确\n- [ ] 语言地道自然（面向全球英文用户）\n- [ ] 标注：目标关键词、漏斗阶段、目标AARRR环节\n- [ ] Meta description ≤155字符，含关键词+行动号召\n- [ ] 图片有alt text（含关键词变体）\n- [ ] URL slug 简短含关键词（英文短横线分隔）\n\nFile v3.0.0:references/agent-refs/data-analyst/methodology.md\n\n# 数据分析方法论\n\n## 反馈分类框架\n\n收到用户反馈后按以下维度分类，再写信给 product-lead：\n\n| 维度 | 分类 | 权重 |\n|------|------|:----:|\n| **类型** | Bug / 功能请求 / 体验抱怨 / 表扬 / 疑问 | — |\n| **频率** | 单次 / 偶发(2-5) / 高频(>5) | 高频×3 |\n| **情绪** | 愤怒 / 失望 / 中性 / 积极 | 愤怒×2 |\n| **用户价值** | 免费 / 付费 / KOL | 付费×2 |\n| **影响范围** | 个人 / 部分用户 / 全体 | 全体×3 |\n\n**写给 product-lead 必须包含：**\n1. 原始反馈（引用原文）\n2. 分类标签\n3. 同类反馈数量（\"过去7天X条类似\"）\n4. 你的分析（真需求还是个例？有 workaround？）\n\n**绝不直接写\"建议开发XX\"** — 你只提供数据+洞察，方案是 product-lead 的事。\n\n## 数据采集清单（按产品生命周期）\n\n### 冷启动期（上线前/刚上线，当前阶段）\n优先接入，成本从低到高：\n\n| 数据源 | 追踪指标 | 工具 | 成本 | 优先级 |\n|--------|---------|------|------|:------:|\n| 搜索引擎 | 收录数、排名、曝光、点击 | Google Search Console + Bing Webmaster | 免费 | P0 |\n| 网站流量 | UV/PV、来源、页面停留、跳出率 | Umami / Plausible (自部署) 或 GA4 | 免费 | P0 |\n| 用户行为 | 点击热图、滚动深度、关键路径 | Microsoft Clarity | 免费 | P1 |\n| 错误监控 | JS报错、API错误率、慢查询 | Sentry (免费额度) | 免费 | P1 |\n| 社媒提及 | 品牌词、竞品词、行业词 | 手动 + intel-analyst | 免费 | P2 |\n| 用户反馈 | 评分、评论、邮件 | 内置反馈widget / G2 | 免费 | P2 |\n\n### 增长期（有稳定流量后）\n| 数据源 | 追踪指标 | 工具 |\n|--------|---------|------|\n| 转化漏斗 | 注册转化率、激活率、付费转化 | 自建事件追踪 |\n| 留存 | D1/D7/D30 留存率 | 数据库查询 |\n| 收入 | ARPU、MRR、Churn | Stripe Dashboard |\n| A/B测试 | 变体胜率、置信度 | GrowthBook (免费) |\n\n### 核心指标仪表盘（上线后必建）\n```\n北极星指标: {待定义，如\"周活跃探索次数\"}\n├── 获客: 日新增UV, 搜索流量占比, 社媒引荐量\n├── 激活: 注册率, 首次搜索完成率, 首次收藏率\n├── 留存: D1/D7 回访率\n├── 收入: (暂缓)\n└── 推荐: 分享率, 外链引用数\n```\n\n## 报告模板\n\n### 周报模板\n```markdown\n# 数据周报 {YYYY}-W{nn}\n\n## 概览\n- 报告周期: {起止日期}\n- 数据源: {列出所有使用的数据源}\n- 数据完整性: ✅完整 / ⚠️部分缺失（注明）/ ❌无数据\n\n## 核心指标\n| 指标 | 本周 | 上周 | 环比 | 趋势 |\n|------|------|------|------|------|\n| UV | | | | ↑↓→ |\n| 注册数 | | | | |\n| 搜索次数 | | | | |\n\n## 流量来源分布\n| 来源 | 占比 | 变化 |\n|------|------|------|\n\n## 异常与预警\n- {异常描述 + 可能原因 + 建议关注}\n\n## 洞察（给 product-lead）\n1. {数据发现 → 含义 → 建议方向（不是方案）}\n\n## 下周关注\n- {需要重点追踪的指标或实验}\n```\n\n### 月报（在周报基础上追加）\n```markdown\n## 月度趋势\n- 本月 vs 上月 vs 上上月（3月对比）\n- 关键拐点和归因分析\n\n## 用户反馈月度汇总\n- 按分类框架统计Top 5主题\n- 情绪趋势变化\n\n## 实验结果汇总\n- 本月实验清单 + 结论（配合 growth-lead）\n```\n\n## 归因分析方法\n\n### 流量归因模型（简易版，适合冷启动）\n| 归因方式 | 适用场景 | 做法 |\n|---------|---------|------|\n| **UTM标签** | 所有主动推广链接 | 统一格式: `?utm_source=X&utm_medium=Y&utm_campaign=Z` |\n| **来源页面 Referrer** | 自然流量 | 分析 referrer 域名分布 |\n| **首次/末次触点** | 转化路径 | 记录用户首次来源 + 注册时来源 |\n| **搜索词** | SEO效果 | Search Console 查询报告 |\n\n### UTM命名规范\n```\nutm_source:   google / bing / reddit / hackernews / twitter / direct\nutm_medium:   organic / social / referral / email / community\nutm_campaign: {描述性名称，如 launch-week, reddit-ama, blog-character-maps}\n```\n\n### 内容→转化归因\n```\n内容贡献 = 该内容页作为入口的注册数 / 总注册数\n```\n定期输出\"内容贡献排行\"给 content-chief，指导内容策略。\n\n## 异常检测规则\n- 日活偏离7日均值 >20% → 预警\n- 关键转化率偏离 >15% → 预警\n- 新增差评 >3条/天 → 预警\n- 搜索排名核心词下降 >10位 → 预警（配合 Search Console）\n- 错误率 >1% → 预警（配合 Sentry）\n- 无数据时：标注\"无数据源，跳过\"，不编造\n\n## 数据呈现标准\n- 必须注明时间范围和数据源\n- 同比环比都给，趋势用 ↑↓→ 标注\n- 不编造数据，没有就说没有\n- 低置信度结论标注 `[低置信度]`\n- 图表优先于纯数字（能给表格就给表格）\n- 关键数字加粗，趋势附原因假设\n\nFile v3.0.0:references/agent-refs/fullstack-dev/coding-behavior-fallback.md\n\n# Coding Behavior (Fallback)\n\n> **This file is only needed when the coding-lead skill is NOT loaded.**\n> If coding-lead is loaded, it provides all these rules in more detail. Ignore this file.\n\n### Task Classification\n| Level | Criteria | Action |\n|-------|----------|--------|\n| Simple | Single file, <60 lines | Do directly with read/write/edit/exec |\n| Medium | 2-5 files, clear scope | Spawn Claude Code via ACP |\n| Complex | Architecture change, multi-module | Plan first, then spawn |\n\n### Context Injection\nBefore spawning Claude Code, gather and inject into prompt:\n1. **产品知识文件** (`shared/products/{product}/` — architecture.md, database.md, services.md 等相关文件)\n2. Project info (CLAUDE.md, README, package.json/composer.json)\n3. Coding standards (tech-standards.md)\n4. Past decisions (search memory for related work)\n5. Known pitfalls from memory\n\n### Prompt Structure for Claude Code\n```\n## Project\n- Path: [project dir]\n- Stack: [from docs]\n\n## Coding Standards\n[From tech-standards.md]\n\n## Historical Context\n[From memory]\n\n## Task\n[Description]\n\n## Acceptance Criteria\n- [ ] ...\n\nBefore finishing:\n1. Run linter if available, fix issues\n2. Run tests if available, ensure they pass\n3. Report results in final output\n\nWhen completely finished, run:\nopenclaw system event --text \"Done: [summary]\" --mode now\n```\n\n### Spawn Rules\n- Always set cwd to the project directory, NOT ~/.openclaw/ or workspace-team/\n- Independent tasks can run in parallel (2-3 sessions max)\n- Track via sessions_list\n- Never let coding agents modify files outside the project directory\n\n### QA Isolation (Critical)\n- QA tests must be spawned in a SEPARATE session from implementation\n- QA prompt gets requirements + interface definitions only, NOT implementation code\n\n### Review by Complexity\n- **Simple**: no review, works = done\n- **Medium**: quick check -- success + tests pass + no obvious errors\n- **Complex**: full checklist (logic, security, performance, style, tests)\n\n### Coding Roles (Complex Tasks Only)\n- **Architect**: system design, DB schema, API contracts\n- **Frontend**: UI components, state management\n- **Backend**: API endpoints, business logic\n- **Reviewer**: independent code review\n- **QA**: test writing, edge case analysis\n\nFlow: Research -> Plan -> Architect(spawn) -> Implement(spawn, can parallel) -> Review(spawn) -> Fix -> Record.\nSkip roles that don't apply. Simple/medium: no roles, single spawn.\n\n### Smart Retry (max 3)\n1. Analyze failure\n2. Rewrite prompt\n3. Retry improved\n4. Max 3 attempts → stop, report to chief-of-staff\n\n### Prompt Pattern Library\n- Record successful prompt structures in memory\n- Search memory for similar past tasks before spawning\n\n### Progress Updates\n- On start/completion/error: notify appropriately\n- Kill runaway sessions and report\n\n## Task Tracking\n\nTrack active coding tasks in `<project>/.openclaw/active-tasks.json`:\n- Register each spawned CC session with: id, task, branch, status, startedAt\n- Update on completion/failure\n- Check before spawning to avoid duplicates\n\n## Definition of Done\n\nMedium: CC success + lint pass + tests pass + no unrelated changes + logged in memory\nComplex: all of above + code review + QA tests + UI screenshots if applicable\n\n## UI Screenshot Rule\nIf a task changes visible UI, the completion report must describe visual changes.\n\nFile v3.0.0:references/agent-refs/fullstack-dev/deep-dive-protocol.md\n\n# Deep Dive Scan Protocol\r\n\r\n> Load on demand: only read when a scan request is received.\r\n\r\n## When to Scan\r\n- Product-lead sends a Deep Dive request via inbox\r\n- New product added to shared/products/_index.md\r\n- After major code changes (L4 incremental)\r\n- Periodic health checks (L3, requested by product-lead)\r\n\r\n## Execution Protocol\r\n\r\n1. **Read the request** from inbox: product name, code directory, scan level, focus areas\r\n2. **Enter the project directory** (read-only unless explicitly told to modify code)\r\n3. **Detect tech stack** automatically:\r\n   - `composer.json` → Laravel/PHP\r\n   - `package.json` → Node/React/Vue (check framework field, dependencies)\r\n   - `requirements.txt` / `pyproject.toml` → Python/Django/FastAPI\r\n   - `go.mod` → Go | `Cargo.toml` → Rust\r\n   - Multiple? Note it's a monorepo or full-stack project\r\n4. **Execute scan commands** per stack (see below)\r\n5. **Write knowledge files** to `shared/products/{product}/`\r\n6. **Log the scan** in `shared/products/{product}/changelog.md`\r\n7. **Notify product-lead** via inbox when done\r\n\r\n## Per-Stack Scan Strategies\r\n\r\n### Laravel/PHP\r\n- L0: tree, composer.json, .env.example\r\n- L1: migrations → database.md, route:list → routes.md + api.md, Models → models.md\r\n- L2: Services/Actions → services.md, Middleware/Policies → auth.md, Jobs/Listeners → jobs-events.md, Console/Kernel → scheduled tasks, config/services.php → integrations.md\r\n- L3: grep TODO/FIXME/HACK, test coverage, Dockerfile, CI, security scan\r\n\r\n### React/Vue/Frontend\r\n- L0: tree src/, package.json\r\n- L1: components → frontend.md, router config, state management stores\r\n- L2: API client layer → data-flow.md, form validation, auth/token handling, i18n\r\n\r\n### Python/Django/FastAPI\r\n- L1: models.py → database.md, urls.py → routes.md, serializers → api.md\r\n- L2: views.py → services.md, middleware/permissions → auth.md, celery → jobs-events.md\r\n\r\n### General (any stack)\r\n- git log --oneline -30, find *.md, grep TODO/FIXME/HACK, tree -L 3\r\n- .env.example → config-env.md, Dockerfile/CI → devops.md, test dirs → test-coverage.md\r\n\r\n## Content Quality Standards\r\n\r\nKnowledge files must capture:\r\n1. **Facts** — what exists (tables, routes, classes)\r\n2. **Relationships** — how things connect\r\n3. **Rationale** — why it's built this way\r\n4. **Implicit rules** — undocumented business logic buried in code\r\n5. **Gotchas** — surprises for a new developer\r\n6. **Cross-module coupling** — where changing A silently breaks B\r\n7. **Performance notes** — N+1 queries, missing indexes, bottlenecks\r\n\r\n## Output Format\r\n\r\n```\r\n# {Title} - {Product Name}\r\n> Auto-generated by the dual-dev Deep Dive flow.\r\n> Last scan: YYYY-MM-DD | Level: L{n} | Stack: {stack}\r\n> Code directory: {path}\r\n```\r\n\r\n## L4 Incremental Protocol\r\n\r\n1. `git log --oneline --since={last scan date}`\r\n2. `git diff --stat` → map changed files to knowledge files\r\n3. Read only changed files, update only affected knowledge files\r\n4. Log in changelog.md\r\n\r\n## Large Projects (>500 files)\r\nSpawn Claude Code for the scan: provide scan level, output directory, content standards. Read-only mode.\n\nFile v3.0.0:references/agent-refs/fullstack-dev/methodology.md\n\n# 工程方法论\n\n## 架构决策记录（ADR）\n\n每个重要技术决策必须留记录。\"重要\"的定义：影响 >1 个模块、不可轻易回退、或团队有分歧。\n\n### ADR 模板\n```markdown\n# ADR-{序号}: {决策标题}\n\n- 日期: {YYYY-MM-DD}\n- 状态: 提议 / 采纳 / 废弃 / 替代(被ADR-XX替代)\n- 决策者: product-lead + fullstack-dev\n\n## 背景\n{为什么需要做这个决策？当前的问题是什么？}\n\n## 决策\n{我们选择了什么方案？}\n\n## 备选方案\n| 方案 | 优点 | 缺点 | 排除原因 |\n|------|------|------|---------|\n| A | | | |\n| B(采纳) | | | — |\n| C | | | |\n\n## 影响\n- 影响模块: {列出}\n- 迁移成本: 高/中/低/无\n- 风险: {可能的问题}\n- 技术债: {引入了什么债/减少了什么债}\n\n## 后续\n- [ ] {需要做的事}\n```\n\n### ADR 存放位置\n`shared/products/{产品}/adrs/` 目录，或写在 architecture.md 的 ## 决策记录 章节。\n\n### 触发 ADR 的场景\n- 新增/替换框架或主要依赖\n- 数据库 schema 重大变更\n- API 破坏性变更\n- 新增微服务/模块拆分\n- 认证/授权方案变更\n- 部署架构变更\n\n## 代码 Review 清单\n\n### 通用检查（所有PR）\n- [ ] **功能正确**: 满足验收标准，边界情况处理\n- [ ] **无回退**: 不破坏已有功能（跑现有测试）\n- [ ] **命名清晰**: 变量/函数/类名可读、一致\n- [ ] **无硬编码**: 配置项走 env/config，不写死\n- [ ] **错误处理**: try-catch 合理，用户友好错误提示\n- [ ] **日志合理**: 关键操作有日志，无敏感信息泄露\n\n### 安全检查\n- [ ] **输入验证**: 所有用户输入已校验/转义\n- [ ] **SQL注入**: 使用参数化查询/ORM，无拼接SQL\n- [ ] **XSS**: 输出已转义，dangerouslySetInnerHTML 有理由\n- [ ] **CSRF**: 表单有token\n- [ ] **认证/授权**: 路由权限检查，无越权访问\n- [ ] **敏感数据**: 密码哈希、token不明文、日志无敏感信息\n- [ ] **依赖安全**: 无已知漏洞依赖（npm audit / composer audit）\n\n### 性能检查\n- [ ] **N+1查询**: 列表页是否 eager load\n- [ ] **索引**: 新增查询条件是否有对应索引\n- [ ] **大数据**: 分页是否正确，无全表扫描\n- [ ] **缓存**: 高频读取是否命中缓存\n- [ ] **前端**: 组件是否合理 memo/懒加载，bundle size\n\n### 可维护性检查\n- [ ] **单一职责**: 函数/组件不超过 100 行（超了要拆）\n- [ ] **重复代码**: DRY，有无可抽取的公共逻辑\n- [ ] **测试**: 新功能有测试，改动不破坏已有测试\n- [ ] **文档**: 复杂逻辑有注释，API 变更更新文档\n- [ ] **兼容性**: 数据库 migration 可回滚\n\n### Review 严格度（按任务复杂度）\n| 复杂度 | Review 方式 |\n|--------|------------|\n| Simple | 自检通过即可（功能ok + lint pass + tests pass） |\n| Medium | 过通用 + 安全清单 |\n| Complex | 全部清单 + 架构影响评估 + 单独QA |\n\n## 技术债评估框架\n\n### 技术债分类\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **架构债** | 系统结构不合理 | 单体过大、循环依赖、缺中间层 |\n| **代码债** | 代码质量差 | 重复代码、过长函数、magic number |\n| **依赖债** | 过时/不安全依赖 | 大版本落后、弃维库、安全漏洞 |\n| **测试债** | 测试覆盖不足 | 无测试、只有happy path、无E2E |\n| **文档债** | 文档缺失/过时 | API无文档、README过时、部署步骤缺失 |\n| **基础设施债** | DevOps/部署问题 | 手动部署、无CI/CD、无监控 |\n\n### 技术债评分（DORA-lite）\n| 因子 | 说明 | 分值 |\n|------|------|:----:|\n| **影响范围** | 影响多少模块/用户 | 1-5 |\n| **修复难度** | 工时估算 | 1-5（5=最难） |\n| **恶化速度** | 不修会不会越来越糟 | 1-5（5=最快恶化） |\n| **阻塞程度** | 是否阻碍新功能开发 | 1-5（5=严重阻塞） |\n\n**技术债优先级 = 影响范围 × 恶化速度 × 阻塞程度 / 修复难度**\n\n分数越高越优先处理。\n\n### 技术债登记格式\n```markdown\n## TD-{序号}: {标题}\n- 类型: 架构/代码/依赖/测试/文档/基础设施\n- 位置: {文件/模块}\n- 评分: 影响={} 难度={} 恶化={} 阻塞={} → 优先级={}\n- 描述: {具体问题}\n- 修复方案: {怎么修}\n- 预估工时: {}\n- 状态: 登记/计划中/修复中/已完成\n```\n\n### 技术债治理规则\n- **每个sprint分配 20% 时间还债**（product-lead 认可）\n- 新功能不得引入同类型的新债（\"先还再借\"）\n- 每月做一次技术债 review，更新评分和优先级\n- 高恶化速度(≥4)的债立即排期，不等sprint\n\n## Git 工作流\n\n### 分支策略（简化 Git Flow）\n```\nmain          ← 生产分支，只接受 merge\n  └── develop ← 开发主线\n       ├── feature/xxx  ← 功能分支\n       ├── fix/xxx      ← 修复分支\n       └── hotfix/xxx   ← 紧急修复（从main拉，合回main+develop）\n```\n\n### Commit 规范\n```\n<type>(<scope>): <subject>\n\ntype: feat|fix|refactor|docs|style|test|chore|perf\nscope: 模块名（可选）\nsubject: 祈使句，不超过72字符\n```\n\n## 部署清单（上线前）\n- [ ] 所有测试通过\n- [ ] Migration 可回滚\n- [ ] 环境变量已配置\n- [ ] 缓存已清理/预热\n- [ ] 监控告警已设置\n- [ ] 回滚方案已确认\n- [ ] 关键路径手动验证\n\nFile v3.0.0:references/agent-refs/growth-lead/methodology.md\n\n# 增长方法论\n\n## AARRR 海盗指标\n\n| 阶段 | 核心指标 | 当前优先级 |\n|------|---------|:----------:|\n| **A**cquisition 获客 | 新访客、流量来源 | ⭐⭐⭐ |\n| **A**ctivation 激活 | 注册率、首次核心操作完成率 | ⭐⭐ |\n| **R**etention 留存 | D1/D7/D30 留存 | ⭐⭐ |\n| **R**evenue 收入 | ARPU、付费转化 | ⭐（暂缓） |\n| **R**eferral 推荐 | NPS、分享率 | ⭐ |\n\n**当前重心：Acquisition + Activation（冷启动期）**\n\n## ICE 实验评分\n\n| 因子 | 说明 | 范围 |\n|------|------|------|\n| **I**mpact | 成功后对指标影响多大 | 1-10 |\n| **C**onfidence | 证据/经验支撑 | 1-10 |\n| **E**ase | 实施难度（10=最易） | 1-10 |\n\n**ICE = (I+C+E) / 3**\n\n## 实验记录格式\n\n```markdown\n## 实验: {标题}\n- ICE: I=_ C=_ E=_ → 分数: _\n- AARRR环节: {哪个}\n- 假设: 如果{做法}，则{指标}提升{预期}%\n- 结果: {数据} → 成功/失败/待定\n- 教训: {下次改进}\n```\n\n## 零预算策略工具箱\n\n| 策略 | 渠道 | 阶段 | 难度 |\n|------|------|------|------|\n| GEO优化 | AI搜索 | 冷启动 | 中 |\n| 长尾SEO | Google/Bing | 中长期 | 低 |\n| 社区种草 | Reddit/HN/Quora | 冷启动 | 低 |\n| PH Launch | Product Hunt | 里程碑 | 高 |\n| 开源引流 | GitHub | 中期 | 中 |\n| 知识图谱占位 | Wiki/Wikidata/G2 | 早期 | 中 |\n\n## 向 product-lead 提需求的规则\n1. **数据依据**：哪些数据表明需要改动？\n2. **预期影响**：哪个指标提升多少？\n3. **不做的代价**？\n4. **最小验证**：能不改代码先验证吗？\n\n**原则：能不动代码就不动代码。增长优先选内容和运营。**\n\nFile v3.0.0:references/agent-refs/intel-analyst/methodology.md\n\n# 竞品分析方法论\n\n## 竞品情报四维框架\n\n每次扫描竞品，从四个维度收集：\n\n| 维度 | 关注点 | 信息源 |\n|------|--------|--------|\n| **产品动态** | 新功能、下线功能、API变更、技术栈变化 | Changelog、博客、GitHub |\n| **商业动态** | 定价变更、融资、合作、收购、裁员 | 新闻、Crunchbase、LinkedIn |\n| **用户声音** | 好评主题、差评主题、迁移原因、NPS趋势 | G2、Capterra、PH、Reddit |\n| **营销动态** | 渠道策略、内容策略、社媒活跃度、广告投放 | SimilarWeb、社媒、搜索 |\n\n## 威胁/机会评估矩阵\n\n每个重要发现必须评估：\n\n| 评估项 | 选项 |\n|--------|------|\n| **性质** | 🔴 威胁 / 🟢 机会 / 🟡 中性 |\n| **紧迫度** | 立即(本周) / 近期(本月) / 观察(本季) |\n| **影响度** | 高(影响核心业务) / 中(影响部分功能) / 低(边缘) |\n| **建议行动** | 防御(我们也要做) / 差异化(做他们不做的) / 忽略 |\n\n## 竞品档案模板\n\n```markdown\n## {竞品名}\n- 定位: {一句话}\n- 主要用户群: {谁在用}\n- 定价: {免费/付费模式}\n- 核心优势: {为什么用户选它}\n- 核心劣势: {用户最大抱怨}\n- 威胁等级: 🔴高 / 🟡中 / 🟢低\n- 最近动态: {最新1-2条}\n- 我们的差异化: {我们做了什么它没做的}\n```\n\n## 扫描优先级\n- **周一全量扫描**: 所有竞品四维全扫\n- **周三/周五快扫**: Top 3 威胁竞品，只看产品+商业动态\n- **>5个竞品时**: Top 3 深扫，其余浅扫（只查新闻）\n\nArchive v2.0.6: 18 files, 55995 bytes\n\nFiles: README.md (4997b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (14240b), references/soul-templates.md (19800b), scripts/deploy.js (21405b), SKILL.md (26360b), team-builder.json (394b), _meta.json (131b)\n\nFile v2.0.6:SKILL.md\n\n---\r\nname: team-builder\r\ndescription: Deploy a multi-agent SaaS growth team on OpenClaw with shared workspace, real-time spawn dispatch (primary) plus async inbox (backup), cron-scheduled tasks, deep project code scanning (Deep Dive), and optional Telegram integration. Use when building or upgrading multi-agent teams for SaaS/product-matrix work. Supports dual-development tracks by default: `devops` for delivery/deploy/environment/acceptance and `fullstack-dev` for implementation/module deep-dive/claude-only coding execution using direct acpx or existing session continuity. Includes Project Deep Dive capability so shared product knowledge files (DB schema, routes, models, services, auth, integrations, tech debt, etc.) can be generated and consumed efficiently by all agents. Supports customizable team name, agent roles, models, timezone, and Telegram bots.\r\n---\r\n\r\n# Team Builder\r\n\r\nDeploy a reusable multi-agent SaaS/growth team template on OpenClaw in one shot.\r\n\r\n## System Impact & Prerequisites\r\n\r\n> **Read before running.** This skill creates files and modifies system config.\r\n\r\n### What it creates\r\n- A new workspace directory with ~40 files (agent configs, shared knowledge, inboxes, kanban)\r\n- `apply-config.js` -- script that **modifies `~/.openclaw/openclaw.json`** (adds agents, bindings, agentToAgent config). Auto-backs up before writing.\r\n- `create-crons.ps1` / `create-crons.sh` -- scripts that **create cron jobs** via `openclaw cron add`\r\n- After running these scripts you must **restart the gateway** (`openclaw gateway restart`)\r\n\r\n### What it does NOT do automatically\r\n- Does not modify openclaw.json directly -- you run `apply-config.js` yourself\r\n- Does not create cron jobs directly -- you run the cron script yourself\r\n- Does not restart the gateway -- you do that manually\r\n\r\n### Optional: Telegram\r\n- If you provide bot tokens during setup, `apply-config.js` will also add Telegram account configs and bindings\r\n- Requires: Telegram bot tokens from @BotFather, your Telegram user ID\r\n- Requires: network access to Telegram API (proxy configurable)\r\n\r\n### Optional: ACP / Claude Code\r\n- The `fullstack-dev` agent is configured as the implementation-focused Claude coding role\r\n- Current production path is **claude only**\r\n- Preferred execution modes: simple direct, medium Claude ACP `run` or direct acpx, complex work via existing fullstack-dev continuity + context files\r\n- Do not assume IM-bound ACP `session` persistence is available\r\n\r\n### Internal Dispatch Protocol (MANDATORY)\r\n- **Standard agents** (product-lead, growth-lead, intel-analyst, devops, etc.): dispatch via `sessions_spawn(runtime=\"subagent\", mode=\"run\")` — **never include `streamTo` parameter**\r\n- **ACP agents** (fullstack-dev): dispatch via `sessions_spawn(runtime=\"acp\")` — may include `streamTo=\"parent\"`\r\n- Results arrive via auto-announce push; do not poll with `sessions_list` or `subagents list`\r\n- Chief-of-staff is the single group entry point; other agents are spawned internally, not expected to independently monitor group chat\r\n- **Stage-progress rule**: if a spawned worker emits meaningful stage progress, the chief-of-staff should forward a compressed user-facing progress update instead of waiting only for final completion\r\n- See `shared/knowledge/team-workflow.md` 零章 for the canonical dispatch specification\r\n\r\n### Action-First Chief Discipline (MANDATORY)\r\n- Chief-of-staff must default to **execute first, report second**\r\n- When the CEO/user says “开干 / 直接推进 / 不要空话”, the current turn must include at least one real action: spawn, edit, verify, commit, or result collection\r\n- Do not output long planning text when the next concrete action is already clear\r\n- Default user-facing updates should be compressed to:\r\n  1. what has been executed\r\n  2. what result came back\r\n  3. what blocker / decision remains\r\n- Inbox is never the default delay mechanism when online agents can be spawned immediately\r\n\r\n### Credentials involved\r\n- **Telegram bot tokens** (optional) -- stored in openclaw.json, used for agent-to-Telegram binding\r\n- **Model API keys** -- must already be configured in your OpenClaw model providers (not handled by this skill)\r\n\r\n### Recommended\r\n- Review generated `apply-config.js` before running\r\n- Check the backup of openclaw.json after running\r\n- Test with 2-3 agents before enabling all cron jobs\r\n\r\n## Team Architecture\r\n\r\nDefault reference architecture for a SaaS/growth multi-agent team (customizable to 2-10 agents):\r\n\r\n```\r\nCEO\r\n |-- Chief of Staff (dispatch + strategy + efficiency)\r\n |-- Data Analyst (data + user research)\r\n |-- Growth Lead (GEO + SEO + community + social media)\r\n |-- Content Chief (strategy + writing + copywriting + i18n)\r\n |-- Intel Analyst (competitor monitoring + market trends)\r\n |-- Product Lead (product management + tech architecture)\r\n |-- DevOps (delivery / deploy / environment / acceptance)\r\n |-- Fullstack Dev (implementation / module deep dive / ACP coding session)\r\n```\r\n\r\n### Multi-Team Support\r\n\r\nOne OpenClaw instance can run multiple teams:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js                  # default team\r\nnode <skill-dir>/scripts/deploy.js --team alpha      # named team \"alpha\"\r\nnode <skill-dir>/scripts/deploy.js --team beta       # named team \"beta\"\r\n```\r\n\r\nNamed teams use prefixed agent IDs (`alpha-chief-of-staff`, `beta-growth-lead`) to avoid conflicts. Each team gets its own workspace subdirectory.\r\n\r\n### Flexible Team Size\r\n\r\nThe wizard lets you select 2-10 agents from the available roles. Skip roles you don't need. The 8-agent default covers most SaaS scenarios with dual-dev routing, but you can run leaner (3-4 agents) or expand with custom roles.\r\n\r\n### Model Auto-Detection\r\n\r\nThe wizard scans your `openclaw.json` for registered model providers and auto-suggests models by role type:\r\n\r\n| Role Type | Best For | Auto-detect Pattern |\r\n|-----------|----------|-------------------|\r\n| Thinking | Strategic roles (chief, growth, content, product) | /glm-5\\|opus\\|o1\\|deepthink/i |\r\n| Execution | Operational roles (data, intel, fullstack) | /glm-4\\|sonnet\\|gpt-4/i |\r\n| Fast | Lightweight tasks | /flash\\|haiku\\|mini/i |\r\n\r\nYou can always override with manual model IDs.\r\n\r\n## Setup / Config / Scripts\r\n\r\n### Required Inputs\r\n- Team name\r\n- Workspace dir\r\n- Timezone\r\n- Morning brief hour\r\n- Evening brief hour\r\n- Thinking model\r\n- Execution model\r\n- CEO title\r\n\r\n### Optional Inputs\r\n- Telegram user ID\r\n- Telegram bot tokens\r\n- Proxy\r\n- ACP coding agent（给 fullstack-dev 使用）\r\n\r\n### Core Scripts\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js\r\nnode <workspace-dir>/apply-config.js\r\npowershell <workspace-dir>/create-crons.ps1\r\nbash <workspace-dir>/create-crons.sh\r\nopenclaw gateway restart\r\n```\r\n\r\n### Execution Priority\r\n- First: matched execution skill (for coding work, `coding-lead` if loaded)\r\n- Second: agent-role fallback when no matching skill is loaded\r\n- Third: templates/README explain boundaries and ownership only; they should not override matched skills\r\n\r\n### Context File Hygiene\r\n- Active context files live under `<project>/.openclaw/`\r\n- Reuse one context file per active code chain when possible\r\n- Naming pattern: `context-<task-slug>.md`\r\n- Active context file cap per project: **60**\r\n- Context-file lifecycle window per project: **100 total files** across active + archive\r\n- Completed or stale files should be deleted or moved to `.openclaw/archive/`\r\n\r\n### Current Dual-Dev Standard\r\n- fullstack-dev：实现、模块深挖、开发文档、接口文档、Claude coding 执行；默认 coding skill 可采用 `coding-lead`，其中 simple 任务直做，medium 倾向 Claude ACP `run` 或 direct acpx，complex 通过现有会话连续协作 + 上下文文件推进，不把 ACP `session` 持久线程作为正式主路径；context 活跃上限 60、生命周期总窗口 100；并行允许但必须先定义边界，总上限 5 个工作单元\r\n- devops：交付、部署、环境、回归、冒烟、自动QA、发布门禁\r\n- product-lead：澄清、PRD、验收标准，不完整不得派工\r\n- chief-of-staff：路由、裁决、控制 token 浪费\r\n\r\n## Deployment Flow\r\n\r\n### Step 1: Collect Configuration\r\n\r\nAsk the user for these inputs (use defaults if not provided):\r\n\r\n| Parameter | Default | Description |\r\n|-----------|---------|-------------|\r\n| Team name | Alpha Team | Used in all docs and configs |\r\n| Workspace dir | `~/.openclaw/workspace-team` | Shared workspace root |\r\n| Timezone | Asia/Shanghai | For cron schedules |\r\n| Morning brief hour | 8 | Chief's morning report |\r\n| Evening brief hour | 18 | Chief's evening report |\r\n| Thinking model | zai/glm-5 | For strategic roles |\r\n| Execution model | zai/glm-4.7 | For execution roles |\r\n| CEO title | Boss | How agents address the CEO |\r\n\r\nOptional: Telegram user ID, proxy, and 7 bot tokens.\r\n\r\n### Step 2: Run Deploy Script\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js\r\n```\r\n\r\nInteractive -- asks all questions from Step 1, generates the full workspace.\r\n\r\n### Step 2b: Non-interactive / Verify Mode\r\n\r\nPrepare a JSON config file:\r\n\r\n```json\r\n{\r\n  \"teamName\": \"Alpha Team\",\r\n  \"workspaceDir\": \"~/.openclaw/workspace-team\",\r\n  \"timezone\": \"Asia/Shanghai\",\r\n  \"morningHour\": 8,\r\n  \"eveningHour\": 18,\r\n  \"thinkingModel\": \"zai/glm-5\",\r\n  \"executionModel\": \"zai/glm-4.7\",\r\n  \"ceoTitle\": \"Boss\",\r\n  \"roles\": [\"chief-of-staff\",\"data-analyst\",\"growth-lead\",\"content-chief\",\"intel-analyst\",\"product-lead\",\"devops\",\"fullstack-dev\"]\r\n}\r\n```\r\n\r\nRun:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js --config team-builder.json\r\nnode <skill-dir>/scripts/deploy.js --verify --config team-builder.json\r\n```\r\n\r\n`--verify` checks that generated files contain the expected dual-dev model, role ownership, and cron entries.\r\n\r\n### Step 3: Apply Config\r\n\r\n```bash\r\nnode <workspace-dir>/apply-config.js\r\n```\r\n\r\nAdds agents to openclaw.json, preserving existing config.\r\n\r\n### Step 4: Create Cron Jobs\r\n\r\n```bash\r\n# Windows\r\npowershell <workspace-dir>/create-crons.ps1\r\n\r\n# Linux/Mac\r\nbash <workspace-dir>/create-crons.sh\r\n```\r\n\r\n### Step 5: Restart Gateway\r\n\r\n```bash\r\nopenclaw gateway restart\r\n```\r\n\r\n### Step 6: Fill Business Info\r\n\r\nUser must edit:\r\n- `shared/decisions/active.md` -- strategy, priorities\r\n- `shared/products/_index.md` -- products overview (≤5 lines per product: URL, code path, positioning, tech, status). Detailed info goes in each product's `overview.md`.\r\n- `shared/knowledge/competitor-map.md` -- competitor analysis\r\n- `shared/knowledge/tech-standards.md` -- coding standards\r\n\r\n### Step 7: Trigger Deep Dive Scans\r\n\r\nAfter filling in products with code directories, tell product-lead to trigger Deep Dive scans:\r\n1. Product-lead dispatches delivery-oriented scan requests to devops (via `sessions_spawn` if online, inbox as fallback)\r\n2. Devops enters each project directory and generates shared knowledge / deployment-oriented scan outputs\r\n3. Fullstack-dev picks up module-level deep dive or implementation follow-up when needed\r\n4. Product-lead reviews the generated files for completeness and acceptance impact\r\n5. All agents now have deep project understanding for informed decisions\r\n\r\n## Cron Schedule\r\n\r\n| Offset | Agent | Task | Frequency |\r\n|--------|-------|------|-----------|\r\n| H-1 | Data Analyst | Data + user feedback | Daily |\r\n| H-1 | Intel Analyst | Competitor scan | Mon/Wed/Fri |\r\n| H | Chief of Staff | Morning brief (announced) | Daily |\r\n| H+1 | Growth Lead | GEO + SEO + community | Daily |\r\n| H+1 | Content Chief | Weekly content plan | Monday |\r\n| H+2 | DevOps | Delivery / environment / Deep Dive / acceptance | Daily |\r\n| H+10 | Chief of Staff | Evening brief (announced) | Daily |\r\n\r\n(H = morning brief hour)\r\n\r\n## Generated File Structure\r\n\r\n```\r\n<workspace>/\r\n├── AGENTS.md, SOUL.md, USER.md  (auto-injected)\r\n├── apply-config.js, create-crons.ps1/.sh, README.md\r\n├── agents/<8 agent dirs>/       (SOUL.md + MEMORY.md + memory/)\r\n└── shared/\r\n    ├── briefings/, decisions/, inbox/ (v2: with status tracking)\r\n    ├── status/team-dashboard.md     (chief-of-staff maintains, all agents read first)\r\n    ├── data/                        (public data pool, data-analyst writes, all read)\r\n    ├── kanban/, knowledge/\r\n    └── products/\r\n        ├── _index.md                (product matrix overview)\r\n        ├── _template/               (knowledge directory template)\r\n        └── {product}/               (per-product knowledge, up to 20 files)\r\n            ├── overview.md, architecture.md, database.md, api.md, routes.md\r\n            ├── models.md, services.md, frontend.md, auth.md, integrations.md\r\n            ├── jobs-events.md, config-env.md, dependencies.md, devops.md\r\n            ├── test-coverage.md, tech-debt.md, domain-flows.md, data-flow.md\r\n            ├── i18n.md, changelog.md, notes.md\r\n```\r\n\r\n\r\n\r\n## Knowledge Governance\r\n\r\nEach shared knowledge file has a designated owner. Only the owner agent updates it; others read only.\r\n\r\n| File | Owner | Update Trigger |\r\n|------|-------|---------------|\r\n| geo-playbook.md | growth-lead | After GEO experiments/discoveries |\r\n| seo-playbook.md | growth-lead | After SEO experiments |\r\n| competitor-map.md | intel-analyst | After each competitor scan |\r\n| content-guidelines.md | content-chief | After proven writing patterns |\r\n| user-personas.md | data-analyst | After new user insights |\r\n| tech-standards.md | product-lead | After architecture decisions |\r\n\r\n### Update Protocol\r\nWhen updating a knowledge file, the owner must:\r\n1. Add a dated entry at the top: `## [YYYY-MM-DD] <what changed>`\r\n2. Include the reason and data evidence\r\n3. Never delete existing entries without CEO approval (append, don't replace)\r\n\r\n### Chief of Staff Governance\r\nThe chief-of-staff monitors knowledge file health during weekly reviews:\r\n- Are files being updated regularly?\r\n- Any conflicting information between files?\r\n- Any stale entries that should be archived?\r\n\r\n## Self-Evolution Pattern\r\n\r\nAgents improve their own strategies over time through a feedback loop:\r\n\r\n```\r\n1. Execute task (cron or inbox triggered)\r\n2. Collect results (data, metrics, outcomes)\r\n3. Analyze: what worked vs what didn't\r\n4. Update knowledge files with proven strategies (with evidence)\r\n5. Next execution reads updated knowledge → better performance\r\n```\r\n\r\nThis is NOT the agent randomly changing rules. Updates must be:\r\n- **Data-driven**: backed by metrics or concrete outcomes\r\n- **Incremental**: append new findings, don't rewrite everything\r\n- **Traceable**: dated with evidence so others can verify\r\n\r\n### What Agents Can Self-Update\r\n- Their own knowledge files (per ownership table above)\r\n- Their own MEMORY.md (lessons learned, decisions)\r\n- shared/data/ outputs (data-analyst only)\r\n\r\n### What Requires CEO Approval\r\n- shared/decisions/active.md (strategy changes)\r\n- Adding/removing agents or changing team architecture\r\n- External publishing or spending decisions\r\n\r\n## Public Data Layer\r\n\r\nThe `shared/data/` directory serves as a read-only data pool for all agents:\r\n\r\n- **data-analyst** writes: daily metrics, user feedback summaries, anomaly alerts\r\n- **All agents** read: to inform their own decisions\r\n- Format: structured markdown or JSON, dated filenames (e.g., `metrics-2026-03-01.md`)\r\n- Retention: keep 30 days, archive older files\r\n\r\n## Project Deep Dive — Code Scanning\r\n\r\nAgents can deeply understand each SaaS product through automated code scanning. This is critical — without deep project knowledge, all team decisions are surface-level.\r\n\r\n### How It Works\r\n\r\n1. CEO adds a product to `shared/products/_index.md` (name, URL, code directory, tech stack)\r\n2. Product Lead triggers a delivery-oriented Deep Dive scan by dispatching to DevOps (via `sessions_spawn` if online, inbox as fallback)\r\n3. DevOps enters the project directory (read-only) and generates shared knowledge / delivery-oriented scan outputs\r\n4. Fullstack Dev picks up module-level deep dive or implementation follow-up when needed\r\n5. Knowledge files are generated in `shared/products/{product}/`\r\n6. All agents consume these files **via manifest-based lazy loading** (never read all at once)\r\n\r\n### Manifest-Based Lazy Loading (MANDATORY)\r\n\r\nEach product directory includes a `manifest.json` (~200 tokens) that lists all files with one-line summaries and a `taskFileMap` mapping task types to relevant files.\r\n\r\n**Agent workflow:**\r\n1. Read `_index.md` → identify which product\r\n2. Read `{product}/manifest.json` → see all files + summaries (~200 tokens)\r\n3. Based on `taskFileMap` or summaries, read only 1-3 relevant files\r\n4. Never read more than 5 product files per session\r\n\r\n**Why:** With 15+ products × 20 files each, full loading = 40K+ tokens per product. Manifest loading = 200 tokens + only what's needed.\r\n\r\n**DevOps MUST regenerate `manifest.json`** after every delivery-oriented scan (L0-L4). Fullstack Dev updates it when doing module-level follow-up that changes knowledge scope. Template in `_template/manifest.json`.\r\n\r\n### Manifest Quality Standards\r\n\r\n摘要不能为了省 token 丢掉关键信息。每条摘要须满足：\r\n- **核心文件**（database/models/services/routes/integrations）：50-130字，列出关键实体名/数量/域名\r\n- **中等文件**（auth/frontend/commands/config）：30-80字，点明方案和范围\r\n- **轻量文件**（changelog/notes/metrics）：可以短（<20字）\r\n- **taskFileMap**：必须覆盖该产品的所有核心业务场景（不少于8个映射）\r\n- **codeStats**：必须包含文件数、行数、模型数、表数等量化指标\r\n\r\n### Product Knowledge Directory\r\n\r\nEach product gets a knowledge directory with up to 20 files + manifest:\r\n\r\n```\r\nshared/products/{product}/\r\n├── manifest.json        ← **INDEX** (~200 tokens): file list, summaries, taskFileMap\r\n├── overview.md          ← Product positioning (from _index.md)\r\n├── architecture.md      ← System architecture, tech stack, design patterns, layering\r\n├── database.md          ← Full table schema, relationships, indexes, migrations\r\n├── api.md               ← API endpoints, params, auth, versioning\r\n├── routes.md            ← Complete route table (Web + API + Console)\r\n├── models.md            ← ORM relationships, scopes, accessors, observers\r\n├── services.md          ← Business logic, state machines, workflows, validation\r\n├── frontend.md          ← Component tree, page routing, state management\r\n├── auth.md              ← Auth scheme, roles/permissions matrix, OAuth\r\n├── integrations.md      ← Third-party: payment/email/SMS/storage/CDN/analytics\r\n├── jobs-events.md       ← Queue jobs, event listeners, scheduled tasks, notifications\r\n├── config-env.md        ← Environment variables, feature flags, cache strategy\r\n├── dependencies.md      ← Key dependencies, custom packages, vulnerabilities\r\n├── devops.md            ← Deployment, CI/CD, Docker, monitoring, logging\r\n├── test-coverage.md     ← Test strategy, coverage, weak spots\r\n├── tech-debt.md         ← TODO/FIXME/HACK inventory, dead code, complexity hotspots\r\n├── domain-flows.md      ← Core user journeys, domain boundaries, module coupling\r\n├── data-flow.md         ← Data lifecycle: external → import → process → store → output\r\n├── i18n.md              ← Internationalization, language coverage\r\n├── changelog.md         ← Scan diff log (what changed between scans)\r\n└── notes.md             ← Agent discoveries, gotchas, implicit rules\r\n```\r\n\r\n### Scan Levels\r\n\r\n| Level | Scope | When | Output |\r\n|-------|-------|------|--------|\r\n| L0 Snapshot | Surface: directory tree, packages, env | First onboard | architecture, dependencies, config-env |\r\n| L1 Skeleton | Structure: DB, routes, models, components | First onboard | database, routes, api, models, frontend |\r\n| L2 Deep Dive | Logic: services, auth, jobs, integrations | On-demand per module | services, auth, jobs-events, integrations, domain-flows, data-flow |\r\n| L3 Health Check | Quality: tech debt, tests, security | Periodic / pre-release | tech-debt, test-coverage, devops |\r\n| L4 Incremental | Delta: git diff → update affected files | After code changes | changelog + targeted updates |\r\n\r\n### Content Standards\r\n\r\nKnowledge files capture not just WHAT exists but WHY:\r\n- **Design decisions**: Why this approach was chosen\r\n- **Implicit business rules**: Logic buried in code (e.g., \"orders auto-cancel after 72h\")\r\n- **Gotchas**: What breaks if you touch this module carelessly\r\n- **Cross-module coupling**: Where changing A silently breaks B\r\n- **Performance hotspots**: N+1 queries, missing indexes, bottleneck endpoints\r\n\r\n### Role Responsibilities\r\n\r\n| Role | Responsibility |\r\n|------|---------------|\r\n| Product Lead | **Clarification / PRD / acceptance**: complete clarification, PRD, user stories, acceptance criteria, and review knowledge freshness before delegating |\r\n| DevOps | **Delivery / QA gate / Deep Dive**: enter code directory for deployment-oriented scans, maintain release checklist, smoke/regression testing, auto-QA access, and generate/update shared product knowledge files |\r\n| Fullstack Dev | **Implementation / docs / Deep Dive follow-up**: continue module-level deep dive, code analysis, implementation, dev docs, interface docs, and ACP session work |\r\n| Chief of Staff | **Routing / escalation**: split implementation vs delivery tasks, prevent duplicate labor, escalate blockers |\r\n| All Agents | **Consumption**: read product knowledge before any product-related decision |\r\n\r\n### Per-Stack Auto-Detection\r\n\r\nFullstack Dev auto-detects tech stack and applies stack-specific scan strategies:\r\n- **Laravel/PHP**: migrations, route:list, Models, Services, Middleware, Policies, Jobs, Console/Kernel\r\n- **React/Vue**: components, router, stores, API client, i18n\r\n- **Python/Django/FastAPI**: models.py, urls.py, views.py, middleware, celery\r\n- **General**: tree, git log, grep TODO/FIXME, .env.example, Docker, CI, tests\r\n\r\n## Team Coordination v2\r\n\r\n### Inbox Protocol v2 (backup channel, status tracking)\r\n\r\n> **Primary dispatch**: `sessions_spawn` (real-time). Inbox is for archival, cross-session handoff, and fallback when spawn is unavailable.\r\n\r\nEvery inbox message now has a `status` field:\r\n- `pending` → `received` → `in-progress` → `done` (or `blocked`)\r\n- Chief-of-staff monitors timeouts: high>4h, normal>24h pending = intervention\r\n- Blocked >8h = escalation to CEO\r\n- Recipients MUST update status immediately upon reading\r\n\r\n### Team Dashboard (`shared/status/team-dashboard.md`)\r\n\r\nChief-of-staff maintains a \"live scoreboard\" updated every session:\r\n- 🔴 Urgent/Blocked items\r\n- 📊 Per-agent status table (last active, current task, status icon)\r\n- 📬 Unprocessed inbox summary (pending/blocked messages across all inboxes)\r\n- 🔗 Cross-agent task chain tracking (A→B→C with per-step status)\r\n- 📅 Today/Tomorrow focus\r\n\r\n**All agents read this file first when waking up.** 5-second situational awareness.\r\n\r\n### Chief-of-Staff as Router\r\n\r\nThe chief is upgraded from \"briefing writer\" to \"active team router\":\r\n- **Real-time dispatch**: uses `sessions_spawn(runtime=\"subagent\")` to directly wake agents and assign tasks — this is the primary dispatch method\r\n- **Blocker detection**: scans all inboxes for overdue messages\r\n- **Inbox as backup**: writes to inbox only for archival, cross-session handoff, or when agent is unreachable\r\n- **Task chain tracking**: identifies multi-agent workflows and tracks each step\r\n- **Escalation**: persistent blockers get flagged to CEO\r\n- **Runs 4x/day** (morning brief, midday patrol, afternoon patrol, evening brief)\r\n\r\n### Cron Schedule (10 jobs, up from 7)\r\n\r\n| Time | Agent | Type | Purpose |\r\n|------|-------|------|---------|\r\n| 07:00 | data-analyst | daily | Data pull + feedback scan |\r\n| 08:00 | chief-of-staff | **announce** | Morning: router scan + brief + quality |\r\n| 09:00 | growth-lead | daily | GEO/SEO/community |\r\n| 09:00 | product-lead | **daily (NEW)** | Inbox + clarification/PRD + task delegation |\r\n| 10:00 | content-chief | **daily M-F (was weekly)** | Content creation + collaboration |\r\n| 10:00 | devops | **daily (delivery track)** | Inbox + Deep Dive + delivery + QA gate |\r\n| 12:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 15:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 18:00 | chief-of-staff | **announce** | Evening: router scan + summary + next day plan |\r\n| 07:00 M/W/F | intel-analyst | 3x/week | Competitor scan |\r\n\r\n### Why These Changes Matter\r\n\r\n| Before | After | Impact |\r\n|--------|-------|--------|\r\n| Inbox = primary dispatch | Inbox = backup + spawn = primary | Real-time dispatch via spawn; inbox for archival only |\r\n| Chief 2x/day | Chief 4x/day with router role | Blockers caught within hours, not days |\r\n| Content-chief 1x/week | Daily M-F | Actually produces content |\r\n| Product-lead no cron | Daily | Knowledge governance happens |\r\n| No team dashboard | Dashboard every session | All agents know the full picture |\r\n| No timeout detection | Automatic timeout rules | Nothing falls through cracks |\r\n\r\n## Key Design Decisions\r\n\r\n- **Shared workspace** so qmd indexes everything for all agents\r\n- **Real-time spawn dispatch** as primary inter-agent communication; inbox as backup for archival and cross-session handoff\r\n- **Chief as Router** — active coordinator who dispatches via `sessions_spawn`, detects blockers, and resolves them\r\n- **Team Dashboard** — single source of truth for team-wide status, maintained by chief every session\r\n- **GEO as #1 priority** (AI search = blue ocean)\r\n- **Fullstack Dev spawns Claude Code** via ACP for complex implementation tasks\r\n- **DevOps owns delivery and QA gate** so implementation and release responsibilities stay separated\r\n- **Project Deep Dive** gives all agents deep codebase understanding, not just surface-level product overviews\r\n\r\n## Customization\r\n\r\nEdit ROLES array in `scripts/deploy.js` to add/remove agents.\r\nEdit `references/soul-templates.md` for SOUL.md templates.\r\nEdit `references/shared-templates.md` for shared file templates.\n\nFile v2.0.6:README.md\n\n# team-builder\n\n## 这是什么\n`team-builder` 用于在 OpenClaw 上生成一套面向 SaaS / 产品矩阵场景的多 Agent 团队工作区与配套脚本。\n\n它负责输出:\n- 团队工作区目录\n- Agent 基础文件\n- shared 协作文件\n- `apply-config.js`\n- `create-crons.ps1` / `create-crons.sh`\n- 角色与协作模板\n\n当前默认采用一套**可复用的双开发轨参考架构**:\n- `devops`:交付、部署、环境、验收、交付向 Deep Dive\n- `fullstack-dev`:项目内实现、模块深挖、代码链连续开发\n\n注意:这是一类多 agent 团队模板,不要求你的团队必须由 team-builder 创建后才成立。\n\n## 适用范围\n适用于:\n- 新建 OpenClaw 团队工作区\n- 升级现有 SaaS 多 Agent 团队模板\n- 为已经存在的多 agent 团队抽取或对齐一套标准化模板\n- 需要内置信箱协作、dashboard、产品知识目录、Deep Dive 流程的团队\n\n不适用于:\n- 直接替你修改线上环境\n- 自动执行 gateway 重启、cron 创建、openclaw.json 写入(这些由你确认后手动执行生成脚本)\n\n## 核心能力\n\n它识别和生成的是\"此类多 agent 协作团队模板\",不是\"唯一正确的团队来源\"。如果你已经有多 agent 团队,只要结构接近,也可以直接吸收它的工作流规则。\n\n\n### 1. 一次生成团队骨架\n生成共享工作区与角色目录,包括:\n- `AGENTS.md`\n- `SOUL.md`\n- `USER.md`\n- `shared/`\n- `agents/`\n- 配置 / cron 脚本\n\n### 2. 双开发轨模板\n内置:\n- `product-lead` 负责澄清 / PRD / 验收 / 路由\n- `devops` 负责交付 / QA gate / delivery-oriented Deep Dive\n- `fullstack-dev` 负责实现 / 模块级 deep dive follow-up / 编码执行\n\n### 3. Deep Dive 产品知识目录\n生成适合沉淀产品知识的共享目录结构,并支持 manifest 驱动的延迟阅读模式。\n\n### 4. 非交互 + verify\n支持:\n- 交互式部署\n- `--config` 非交互部署\n- `--verify` 校验生成结果\n\n当前 `--verify` 会输出**结构化 JSON 检查报告**,覆盖:\n- core\n- roles\n- fallback\n- workflow\n- hygiene\n- cron\n\n### 5. fallback 与上下文治理同步\n已对齐当前编码执行口径:\n- Claude-only ACP 默认口径\n- context active cap:**60**\n- lifecycle window:**100**\n- done 前 verify\n- 写入 / spawn 前确认 `cwd`\n\n## 初始化\n本技能**无需单独初始化**,但第一次使用前建议确认:\n- OpenClaw 已正确安装\n- Node.js 可用\n- 你有权查看并手动执行生成出来的脚本\n\n## 使用方法\n\n### 1. 交互式生成\n```bash\nnode <skill-dir>/scripts/deploy.js\n```\n适合首次搭团队。\n\n### 2. 非交互生成\n准备一个 JSON 配置文件,再执行:\n```bash\nnode <skill-dir>/scripts/deploy.js --config team-builder.json\n```\n\n### 3. 验证生成结果\n```bash\nnode <skill-dir>/scripts/deploy.js --verify --config team-builder.json\n```\n输出结构化 JSON 检查报告;若失败,进程以非 0 退出。\n\n### 4. 应用生成结果\n生成完成后,按顺序手动执行:\n1. `node <workspace-dir>/apply-config.js`\n2. `powershell <workspace-dir>/create-crons.ps1` 或 `bash <workspace-dir>/create-crons.sh`\n3. `openclaw gateway restart`\n\n## 配置说明\n\n### 必填项\n- teamName\n- workspaceDir\n- timezone\n- morningHour\n- eveningHour\n- thinkingModel\n- executionModel\n- ceoTitle\n\n### 可选项\n- roles\n- roleNames\n- `--team` 前缀\n- Telegram 相关输入(如果你准备接入 Telegram)\n\n## 输出物说明\n\n### 1. 工作区\n输出团队工作区,包含:\n- 角色目录\n- inbox / briefings / dashboard\n- products / knowledge / kanban\n- 团队工作流默认规则(任务入口、done 链、最小读取顺序、角色记忆边界)\n\n### 2. `apply-config.js`\n用于把生成的 Agent 写入 `openclaw.json`。\n**注意:不是自动执行,而是生成后由你手动执行。**\n\n### 3. `create-crons.*`\n用于生成 cron 任务。\n**注意:不是自动执行,而是生成后由你手动执行。**\n\n### 4. references/fallback 文件\n生成的团队工作区内现在会包含:\n- `references/coding-behavior-fallback.md`\n\n供 `fullstack-dev` 在 `coding-lead` 未加载时使用。\n\n## 验证建议\n改动本技能时,至少确认:\n- `deploy.js --verify` 仍输出结构化 JSON 报告\n- 生成物里仍包含 fallback 文件和 team-workflow.md\n- README / SKILL / templates / deploy.js 文案一致\n- context 规则仍是 **60 / 100**\n\n## 相关文件\n- `SKILL.md`\n- `scripts/deploy.js`\n- `references/shared-templates.md`\n- `references/soul-templates.md`\n- `references/coding-behavior-fallback.md`\n\n## 最近一次修改(中文)\n- **2026-03-29 / 7502460**\n- 变更摘要:\n  - `deploy.js --verify` 扩展为**结构化 JSON 检查报告**,报告按 `core / roles / fallback / workflow / hygiene / cron` 分组\n  - 生成物自动带 `shared/knowledge/team-workflow.md`(任务入口、done 链、最小读取顺序、角色记忆边界)\n  - `shared-templates.md` 改为指向独立 workflow 规范文件,不再内嵌细则\n  - 定位从\"部署固定 8-agent 团队\"改为\"生成可复用的此类多 agent 团队模板\"\n\nFile v2.0.6:_meta.json\n\n{\n  \"ownerId\": \"kn761rne5jajzfg0d8h6nge4hh8220cq\",\n  \"slug\": \"team-builder\",\n  \"version\": \"2.0.6\",\n  \"publishedAt\": 1774831815771\n}\n\nFile v2.0.6:references/agent-refs/chief-of-staff/dashboard-template.md\n\n# 团队面板模板\n\n```markdown\n# 团队状态面板\n> 参谋长维护 | 最后更新: YYYY-MM-DD HH:MM\n\n## 🔴 紧急/阻塞\n- [描述] — 涉及: [agent] — 原因: [原因] — 已持续: [时长]\n\n## 📊 各agent状态\n| Agent | 上次活跃 | 当前任务 | 状态 |\n|-------|---------|---------|------|\n| 参谋长 | 刚才 | 面板+简报 | ✅ |\n| 产品总监 | | | |\n| 增长总监 | | | |\n| 内容主编 | | | |\n| 数据分析 | | | |\n| 情报分析 | | | |\n| 全栈工程 | | | |\n\n状态: ✅正常 🔄进行中 🔴阻塞 ⏳等待 💤失联(>48h)\n\n## 📬 未处理信箱\n| 消息 | 发→收 | 优先级 | 已等待 | 状态 |\n\n## 🔗 跨agent任务链\n- [任务]: A(✅) → B(🔄) → C(⏳)\n\n## 📅 今日/明日重点\n1. ...\n```\n\nFile v2.0.6:references/agent-refs/chief-of-staff/strategy-methodology.md\n\n# 矩阵策略方法论\n\n## 产品组合评估框架\n\n### 产品健康度评分卡（每产品，月度更新）\n\n| 维度 | 指标 | 权重 | 评分(1-5) | 说明 |\n|------|------|:----:|:---------:|------|\n| **流量** | 月UV增长率 | 20% | | >20%=5, 10-20%=4, 0-10%=3, 持平=2, 下降=1 |\n| **留存** | D7回访率 | 20% | | >40%=5, 30-40%=4, 20-30%=3, 10-20%=2, <10%=1 |\n| **变现** | 收入/利润趋势 | 15% | | 盈利=5, 接近盈亏=4, 有收入=3, 无收入=2, 亏损=1 |\n| **竞争** | 差异化强度 | 15% | | 独占优势=5, 明显差异=4, 有差异=3, 同质=2, 落后=1 |\n| **效率** | 投入产出比 | 15% | | 高ROI=5, 正向=4, 中等=3, 低=2, 负=1 |\n| **战略** | 对矩阵协同价值 | 15% | | 核心枢纽=5, 高协同=4, 有协同=3, 独立=2, 拖累=1 |\n\n**总分 = Σ(评分 × 权重)**\n\n### 产品分类（简化BCG矩阵）\n\n| 类型 | 条件 | 策略 |\n|------|------|------|\n| ⭐ **明星** | 高增长 + 高份额（或潜力） | 全力投入，抢市场 |\n| 🐄 **现金牛** | 低增长 + 高份额 | 维护为主，收割利润，少投入 |\n| ❓ **问号** | 高增长 + 低份额 | 限期验证：3个月达标→升星，否则降级 |\n| 🐕 **瘦狗** | 低增长 + 低份额 | 最低维护或考虑下线，资源释放 |\n\n### 产品组合仪表盘模板\n```markdown\n# 产品矩阵健康度 — {YYYY}-{MM}\n\n| 产品 | 类型 | 健康度 | 本月重点 | 资源分配 |\n|------|------|:------:|---------|:--------:|\n| Richlyva | ❓问号 | 3.2/5 | 冷启动验证 | 40% |\n| ... | | | | |\n\n## 矩阵级风险\n- {跨产品风险}\n\n## 协同机会\n- {跨产品引流/共享组件}\n```\n\n## 资源分配模型\n\n### 核心原则\nCEO一个人 + AI团队，资源 = **CEO时间 + Agent算力 + 注意力**。\n\n### 分配公式\n```\n产品资源占比 = (健康度得分 × 战略权重) / Σ(所有产品的 健康度×权重)\n```\n\n其中**战略权重**由CEO指定（1-3倍）：\n- 3× = CEO明确标注的战略重点\n- 2× = 有明确增长潜力\n- 1× = 维护型\n\n### 分配约束\n| 规则 | 约束 |\n|------|------|\n| 单产品上限 | ≤60%（避免all-in风险） |\n| 单产品下限 | ≥5%（维护最低限度） |\n| 新产品试验 | 独立预留10-15%（不从现有产品扣） |\n| 调整频率 | 月度review，CEO批准后生效 |\n\n### Agent工时分配（参考）\n| 角色 | 分配逻辑 |\n|------|---------|\n| fullstack-dev | 按产品资源占比分配coding时间 |\n| content-chief | 按产品资源占比分配内容产出 |\n| growth-lead | 聚焦当前AARRR最弱环节的产品 |\n| data-analyst | 各产品均匀覆盖（数据不能偏科） |\n| intel-analyst | 聚焦⭐明星和❓问号的竞品 |\n\n### 资源重分配触发条件\n| 信号 | 动作 |\n|------|------|\n| 产品健康度连续2月下降 | 减少投入，除非CEO坚持 |\n| 产品健康度突破4.0 | 考虑加大投入（明星候选） |\n| ❓问号产品3月未达标 | 降为🐕瘦狗，释放资源 |\n| CEO新增产品 | 从预留池分配，不挤占现有 |\n| 竞品重大威胁 | 临时增加相关产品投入（参谋长判断+CEO批准） |\n\n## 跨产品协同策略\n\n### 协同类型\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **流量协同** | A产品用户引导到B产品 | 影视站→工具站推荐 |\n| **技术协同** | 共享组件/服务 | 统一认证、共享UI库 |\n| **数据协同** | 共享用户画像/行为数据 | 用户兴趣标签跨产品复用 |\n| **品牌协同** | 统一品牌矩阵 | 主品牌→子品牌信任传递 |\n| **SEO协同** | 站群互链+内容交叉引用 | 但要注意搜索引擎惩罚风险 |\n\n### 协同评估\n每次发现协同机会时评估：\n- **收益**: 预期带来多少增量（流量/用户/收入）\n- **成本**: 需要多少开发/运营投入\n- **风险**: 是否有负面影响（如SEO惩罚、品牌稀释）\n- **依赖**: 需要几个Agent配合，协调成本高不高\n\n## 定价策略分析框架\n\n### 定价三角\n```\n        价值感知\n       /        \\\n      /          \\\n竞品价格 ——— 成本底线\n```\n\n### 定价方法选择\n| 阶段 | 方法 | 说明 |\n|------|------|------|\n| 冷启动 | 免费/Freemium | 先要用户，不要收入 |\n| 有留存 | 功能分级 | 基础免费 + 高级付费 |\n| 有品牌 | 价值定价 | 按用户获得的价值定价 |\n| 成熟期 | 竞品对标 | 参考竞品定价+差异化溢价 |\n\n### 价格测试框架\n- A/B测试不同价格点（需要足够流量）\n- 用户调研：\"你愿意为X付多少\"\n- 竞品定价对标分析（intel-analyst 提供数据）\n- Van Westendorp价格敏感度（如果能做调研的话）\n\nFile v2.0.6:references/agent-refs/content-chief/methodology.md\n\n# 内容方法论\n\n## 内容漏斗匹配（Content-Funnel Fit）\n\n每篇内容明确服务哪个用户旅程阶段：\n\n| 阶段 | 用户意图 | 内容类型 | 示例 |\n|------|---------|---------|------|\n| **TOFU** 认知 | \"有这类工具吗？\" | 科普、趋势、问题解答 | \"2026年影视数据库对比\" |\n| **MOFU** 考虑 | \"哪个适合我？\" | 对比评测、教程、案例 | \"Richlyva vs IMDb 角色图谱对比\" |\n| **BOFU** 决策 | \"我要用这个\" | 上手指南、API文档 | \"5分钟搭建影视宇宙\" |\n\n**比例约 5:3:2（TOFU:MOFU:BOFU）**\n\n## 内容日历与排期\n\n### 发布节奏（冷启动期）\n| 类型 | 频率 | 负责 | 说明 |\n|------|------|------|------|\n| 长文（>1500词） | 2篇/周 | 内容主编主导 | SEO/GEO核心资产 |\n| 短文/列表（500-1000词） | 2篇/周 | 可委托或模板化 | 长尾关键词覆盖 |\n| 社区帖（Reddit/HN/Quora） | 3条/周 | 配合growth-lead | 引流+品牌曝光 |\n| FAQ/更新日志 | 按需 | 配合product-lead | 站内SEO+用户信任 |\n\n### 选题流程（4步）\n1. **关键词池** → 从 SEO/GEO 关键词研究中选取本周目标词\n2. **意图匹配** → 判断搜索意图（信息型/导航型/交易型）→ 匹配漏斗阶段\n3. **竞品内容差距** → intel-analyst 竞品数据中找对手没覆盖的角度\n4. **排期** → 写入 `shared/knowledge/content-plan-{YYYY}-W{nn}.md`\n\n### 内容日历模板\n```markdown\n## 内容计划 {YYYY}-W{nn}\n| 日期 | 标题 | 漏斗 | 目标关键词 | 渠道 | 状态 |\n|------|------|------|-----------|------|------|\n| 周一 | ... | TOFU | ... | Blog | 草稿/发布/推迟 |\n```\n\n## SEO/GEO 关键词研究流程\n\n### 第1步：种子词收集\n- 产品核心功能词（character map, movie universe, film database）\n- 用户痛点词（从 data-analyst 反馈/搜索query）\n- 竞品品牌词+替代词（IMDb alternative, letterboxd vs）\n- AI搜索常见问法（best tool for..., how to find...）\n\n### 第2步：扩展与验证\n| 工具/方法 | 用途 | 成本 |\n|-----------|------|------|\n| Google Suggest / People Also Ask | 长尾扩展 | 免费 |\n| Bing Webmaster | 实际搜索数据 | 免费 |\n| 竞品页面H1/H2/Title扫描 | 发现覆盖差距 | 免费 |\n| Reddit/Quora热门问题 | 真实用户措辞 | 免费 |\n| Google Trends | 趋势验证 | 免费 |\n\n### 第3步：关键词评分\n| 因子 | 说明 | 权重 |\n|------|------|:----:|\n| 搜索量预估 | 高/中/低（无精确数据时用趋势判断） | 30% |\n| 竞争难度 | SERP前10是大站还是小站 | 25% |\n| 意图匹配 | 和产品功能的相关度 | 25% |\n| GEO潜力 | AI搜索是否会引用此类内容 | 20% |\n\n### 第4步：关键词档案\n```markdown\n## 关键词: {keyword}\n- 搜索量预估: 高/中/低\n- 竞争难度: 高/中/低\n- 意图: 信息型/导航型/交易型\n- 漏斗阶段: TOFU/MOFU/BOFU\n- SERP现状: {前3结果概述}\n- GEO机会: {AI搜索是否已有引用}\n- 优先级: P1/P2/P3\n- 计划内容: {标题草案}\n```\n\n## 内容效果评估框架\n\n### 发布后追踪（分阶段）\n| 时间点 | 检查项 | 合格线 |\n|--------|--------|--------|\n| 发布即时 | Google索引（site:命令） | 48h内收录 |\n| 7天 | Search Console曝光量 | >0（有曝光） |\n| 14天 | 排名位置 | 目标词进前50 |\n| 30天 | 点击量、平均排名 | 目标词进前20 |\n| 60天 | 自然流量占比 | 该页有稳定流量 |\n\n### 内容健康度分级\n| 等级 | 标准 | 行动 |\n|------|------|------|\n| 🟢 健康 | 排名稳定/上升，有流量 | 保持，考虑扩展内链 |\n| 🟡 待优化 | 排名11-30，有曝光无点击 | 优化标题/摘要/结构 |\n| 🔴 失败 | 60天无排名无流量 | 分析原因：选题错？内容薄？竞争强？ |\n| ⚫ 过时 | >6个月未更新且流量下降 | 更新或合并 |\n\n### 内容ROI评估\n```\n内容价值 = (自然流量 × 预估CPC) / 创作时间(小时)\n```\n无精确数据时用定性判断：这篇内容带来了注册/咨询/品牌曝光吗？\n\n## 内容复用策略（1→N）\n\n一篇核心长文可拆解为：\n\n| 衍生形式 | 渠道 | 改造量 |\n|---------|------|--------|\n| 精华摘要（300字） | Reddit/HN帖子 | 低 |\n| FAQ拆条 | Quora回答 | 低 |\n| 数据可视化 | Twitter/社媒图 | 中 |\n| 对比表格 | 独立落地页 | 中 |\n| 视频脚本大纲 | YouTube（未来） | 高 |\n| Newsletter片段 | 邮件（未来） | 低 |\n\n**规则：核心长文先发博客 → 等收录后 → 再拆分发各渠道（避免重复内容惩罚）**\n\n## 写作检查清单（发布前必检）\n- [ ] 标题含关键词且吸引点击\n- [ ] 开头30字回答搜索意图（AI搜索截取用）\n- [ ] H2/H3 含关键词变体\n- [ ] FAQ 区块 3-5 个（Schema+GEO友好）\n- [ ] 内链 ≥2 个\n- [ ] CTA 明确\n- [ ] 语言地道自然（面向全球英文用户）\n- [ ] 标注：目标关键词、漏斗阶段、目标AARRR环节\n- [ ] Meta description ≤155字符，含关键词+行动号召\n- [ ] 图片有alt text（含关键词变体）\n- [ ] URL slug 简短含关键词（英文短横线分隔）\n\nFile v2.0.6:references/agent-refs/data-analyst/methodology.md\n\n# 数据分析方法论\n\n## 反馈分类框架\n\n收到用户反馈后按以下维度分类，再写信给 product-lead：\n\n| 维度 | 分类 | 权重 |\n|------|------|:----:|\n| **类型** | Bug / 功能请求 / 体验抱怨 / 表扬 / 疑问 | — |\n| **频率** | 单次 / 偶发(2-5) / 高频(>5) | 高频×3 |\n| **情绪** | 愤怒 / 失望 / 中性 / 积极 | 愤怒×2 |\n| **用户价值** | 免费 / 付费 / KOL | 付费×2 |\n| **影响范围** | 个人 / 部分用户 / 全体 | 全体×3 |\n\n**写给 product-lead 必须包含：**\n1. 原始反馈（引用原文）\n2. 分类标签\n3. 同类反馈数量（\"过去7天X条类似\"）\n4. 你的分析（真需求还是个例？有 workaround？）\n\n**绝不直接写\"建议开发XX\"** — 你只提供数据+洞察，方案是 product-lead 的事。\n\n## 数据采集清单（按产品生命周期）\n\n### 冷启动期（上线前/刚上线，当前阶段）\n优先接入，成本从低到高：\n\n| 数据源 | 追踪指标 | 工具 | 成本 | 优先级 |\n|--------|---------|------|------|:------:|\n| 搜索引擎 | 收录数、排名、曝光、点击 | Google Search Console + Bing Webmaster | 免费 | P0 |\n| 网站流量 | UV/PV、来源、页面停留、跳出率 | Umami / Plausible (自部署) 或 GA4 | 免费 | P0 |\n| 用户行为 | 点击热图、滚动深度、关键路径 | Microsoft Clarity | 免费 | P1 |\n| 错误监控 | JS报错、API错误率、慢查询 | Sentry (免费额度) | 免费 | P1 |\n| 社媒提及 | 品牌词、竞品词、行业词 | 手动 + intel-analyst | 免费 | P2 |\n| 用户反馈 | 评分、评论、邮件 | 内置反馈widget / G2 | 免费 | P2 |\n\n### 增长期（有稳定流量后）\n| 数据源 | 追踪指标 | 工具 |\n|--------|---------|------|\n| 转化漏斗 | 注册转化率、激活率、付费转化 | 自建事件追踪 |\n| 留存 | D1/D7/D30 留存率 | 数据库查询 |\n| 收入 | ARPU、MRR、Churn | Stripe Dashboard |\n| A/B测试 | 变体胜率、置信度 | GrowthBook (免费) |\n\n### 核心指标仪表盘（上线后必建）\n```\n北极星指标: {待定义，如\"周活跃探索次数\"}\n├── 获客: 日新增UV, 搜索流量占比, 社媒引荐量\n├── 激活: 注册率, 首次搜索完成率, 首次收藏率\n├── 留存: D1/D7 回访率\n├── 收入: (暂缓)\n└── 推荐: 分享率, 外链引用数\n```\n\n## 报告模板\n\n### 周报模板\n```markdown\n# 数据周报 {YYYY}-W{nn}\n\n## 概览\n- 报告周期: {起止日期}\n- 数据源: {列出所有使用的数据源}\n- 数据完整性: ✅完整 / ⚠️部分缺失（注明）/ ❌无数据\n\n## 核心指标\n| 指标 | 本周 | 上周 | 环比 | 趋势 |\n|------|------|------|------|------|\n| UV | | | | ↑↓→ |\n| 注册数 | | | | |\n| 搜索次数 | | | | |\n\n## 流量来源分布\n| 来源 | 占比 | 变化 |\n|------|------|------|\n\n## 异常与预警\n- {异常描述 + 可能原因 + 建议关注}\n\n## 洞察（给 product-lead）\n1. {数据发现 → 含义 → 建议方向（不是方案）}\n\n## 下周关注\n- {需要重点追踪的指标或实验}\n```\n\n### 月报（在周报基础上追加）\n```markdown\n## 月度趋势\n- 本月 vs 上月 vs 上上月（3月对比）\n- 关键拐点和归因分析\n\n## 用户反馈月度汇总\n- 按分类框架统计Top 5主题\n- 情绪趋势变化\n\n## 实验结果汇总\n- 本月实验清单 + 结论（配合 growth-lead）\n```\n\n## 归因分析方法\n\n### 流量归因模型（简易版，适合冷启动）\n| 归因方式 | 适用场景 | 做法 |\n|---------|---------|------|\n| **UTM标签** | 所有主动推广链接 | 统一格式: `?utm_source=X&utm_medium=Y&utm_campaign=Z` |\n| **来源页面 Referrer** | 自然流量 | 分析 referrer 域名分布 |\n| **首次/末次触点** | 转化路径 | 记录用户首次来源 + 注册时来源 |\n| **搜索词** | SEO效果 | Search Console 查询报告 |\n\n### UTM命名规范\n```\nutm_source:   google / bing / reddit / hackernews / twitter / direct\nutm_medium:   organic / social / referral / email / community\nutm_campaign: {描述性名称，如 launch-week, reddit-ama, blog-character-maps}\n```\n\n### 内容→转化归因\n```\n内容贡献 = 该内容页作为入口的注册数 / 总注册数\n```\n定期输出\"内容贡献排行\"给 content-chief，指导内容策略。\n\n## 异常检测规则\n- 日活偏离7日均值 >20% → 预警\n- 关键转化率偏离 >15% → 预警\n- 新增差评 >3条/天 → 预警\n- 搜索排名核心词下降 >10位 → 预警（配合 Search Console）\n- 错误率 >1% → 预警（配合 Sentry）\n- 无数据时：标注\"无数据源，跳过\"，不编造\n\n## 数据呈现标准\n- 必须注明时间范围和数据源\n- 同比环比都给，趋势用 ↑↓→ 标注\n- 不编造数据，没有就说没有\n- 低置信度结论标注 `[低置信度]`\n- 图表优先于纯数字（能给表格就给表格）\n- 关键数字加粗，趋势附原因假设\n\nFile v2.0.6:references/agent-refs/fullstack-dev/coding-behavior-fallback.md\n\n# Coding Behavior (Fallback)\n\n> **This file is only needed when the coding-lead skill is NOT loaded.**\n> If coding-lead is loaded, it provides all these rules in more detail. Ignore this file.\n\n### Task Classification\n| Level | Criteria | Action |\n|-------|----------|--------|\n| Simple | Single file, <60 lines | Do directly with read/write/edit/exec |\n| Medium | 2-5 files, clear scope | Spawn Claude Code via ACP |\n| Complex | Architecture change, multi-module | Plan first, then spawn |\n\n### Context Injection\nBefore spawning Claude Code, gather and inject into prompt:\n1. **产品知识文件** (`shared/products/{product}/` — architecture.md, database.md, services.md 等相关文件)\n2. Project info (CLAUDE.md, README, package.json/composer.json)\n3. Coding standards (tech-standards.md)\n4. Past decisions (search memory for related work)\n5. Known pitfalls from memory\n\n### Prompt Structure for Claude Code\n```\n## Project\n- Path: [project dir]\n- Stack: [from docs]\n\n## Coding Standards\n[From tech-standards.md]\n\n## Historical Context\n[From memory]\n\n## Task\n[Description]\n\n## Acceptance Criteria\n- [ ] ...\n\nBefore finishing:\n1. Run linter if available, fix issues\n2. Run tests if available, ensure they pass\n3. Report results in final output\n\nWhen completely finished, run:\nopenclaw system event --text \"Done: [summary]\" --mode now\n```\n\n### Spawn Rules\n- Always set cwd to the project directory, NOT ~/.openclaw/ or workspace-team/\n- Independent tasks can run in parallel (2-3 sessions max)\n- Track via sessions_list\n- Never let coding agents modify files outside the project directory\n\n### QA Isolation (Critical)\n- QA tests must be spawned in a SEPARATE session from implementation\n- QA prompt gets requirements + interface definitions only, NOT implementation code\n\n### Review by Complexity\n- **Simple**: no review, works = done\n- **Medium**: quick check -- success + tests pass + no obvious errors\n- **Complex**: full checklist (logic, security, performance, style, tests)\n\n### Coding Roles (Complex Tasks Only)\n- **Architect**: system design, DB schema, API contracts\n- **Frontend**: UI components, state management\n- **Backend**: API endpoints, business logic\n- **Reviewer**: independent code review\n- **QA**: test writing, edge case analysis\n\nFlow: Research -> Plan -> Architect(spawn) -> Implement(spawn, can parallel) -> Review(spawn) -> Fix -> Record.\nSkip roles that don't apply. Simple/medium: no roles, single spawn.\n\n### Smart Retry (max 3)\n1. Analyze failure\n2. Rewrite prompt\n3. Retry improved\n4. Max 3 attempts → stop, report to chief-of-staff\n\n### Prompt Pattern Library\n- Record successful prompt structures in memory\n- Search memory for similar past tasks before spawning\n\n### Progress Updates\n- On start/completion/error: notify appropriately\n- Kill runaway sessions and report\n\n## Task Tracking\n\nTrack active coding tasks in `<project>/.openclaw/active-tasks.json`:\n- Register each spawned CC session with: id, task, branch, status, startedAt\n- Update on completion/failure\n- Check before spawning to avoid duplicates\n\n## Definition of Done\n\nMedium: CC success + lint pass + tests pass + no unrelated changes + logged in memory\nComplex: all of above + code review + QA tests + UI screenshots if applicable\n\n## UI Screenshot Rule\nIf a task changes visible UI, the completion report must describe visual changes.\n\nFile v2.0.6:references/agent-refs/fullstack-dev/deep-dive-protocol.md\n\n# Deep Dive Scan Protocol\r\n\r\n> Load on demand: only read when a scan request is received.\r\n\r\n## When to Scan\r\n- Product-lead sends a Deep Dive request via inbox\r\n- New product added to shared/products/_index.md\r\n- After major code changes (L4 incremental)\r\n- Periodic health checks (L3, requested by product-lead)\r\n\r\n## Execution Protocol\r\n\r\n1. **Read the request** from inbox: product name, code directory, scan level, focus areas\r\n2. **Enter the project directory** (read-only unless explicitly told to modify code)\r\n3. **Detect tech stack** automatically:\r\n   - `composer.json` → Laravel/PHP\r\n   - `package.json` → Node/React/Vue (check framework field, dependencies)\r\n   - `requirements.txt` / `pyproject.toml` → Python/Django/FastAPI\r\n   - `go.mod` → Go | `Cargo.toml` → Rust\r\n   - Multiple? Note it's a monorepo or full-stack project\r\n4. **Execute scan commands** per stack (see below)\r\n5. **Write knowledge files** to `shared/products/{product}/`\r\n6. **Log the scan** in `shared/products/{product}/changelog.md`\r\n7. **Notify product-lead** via inbox when done\r\n\r\n## Per-Stack Scan Strategies\r\n\r\n### Laravel/PHP\r\n- L0: tree, composer.json, .env.example\r\n- L1: migrations → database.md, route:list → routes.md + api.md, Models → models.md\r\n- L2: Services/Actions → services.md, Middleware/Policies → auth.md, Jobs/Listeners → jobs-events.md, Console/Kernel → scheduled tasks, config/services.php → integrations.md\r\n- L3: grep TODO/FIXME/HACK, test coverage, Dockerfile, CI, security scan\r\n\r\n### React/Vue/Frontend\r\n- L0: tree src/, package.json\r\n- L1: components → frontend.md, router config, state management stores\r\n- L2: API client layer → data-flow.md, form validation, auth/token handling, i18n\r\n\r\n### Python/Django/FastAPI\r\n- L1: models.py → database.md, urls.py → routes.md, serializers → api.md\r\n- L2: views.py → services.md, middleware/permissions → auth.md, celery → jobs-events.md\r\n\r\n### General (any stack)\r\n- git log --oneline -30, find *.md, grep TODO/FIXME/HACK, tree -L 3\r\n- .env.example → config-env.md, Dockerfile/CI → devops.md, test dirs → test-coverage.md\r\n\r\n## Content Quality Standards\r\n\r\nKnowledge files must capture:\r\n1. **Facts** — what exists (tables, routes, classes)\r\n2. **Relationships** — how things connect\r\n3. **Rationale** — why it's built this way\r\n4. **Implicit rules** — undocumented business logic buried in code\r\n5. **Gotchas** — surprises for a new developer\r\n6. **Cross-module coupling** — where changing A silently breaks B\r\n7. **Performance notes** — N+1 queries, missing indexes, bottlenecks\r\n\r\n## Output Format\r\n\r\n```\r\n# {Title} - {Product Name}\r\n> Auto-generated by the dual-dev Deep Dive flow.\r\n> Last scan: YYYY-MM-DD | Level: L{n} | Stack: {stack}\r\n> Code directory: {path}\r\n```\r\n\r\n## L4 Incremental Protocol\r\n\r\n1. `git log --oneline --since={last scan date}`\r\n2. `git diff --stat` → map changed files to knowledge files\r\n3. Read only changed files, update only affected knowledge files\r\n4. Log in changelog.md\r\n\r\n## Large Projects (>500 files)\r\nSpawn Claude Code for the scan: provide scan level, output directory, content standards. Read-only mode.\n\nFile v2.0.6:references/agent-refs/fullstack-dev/methodology.md\n\n# 工程方法论\n\n## 架构决策记录（ADR）\n\n每个重要技术决策必须留记录。\"重要\"的定义：影响 >1 个模块、不可轻易回退、或团队有分歧。\n\n### ADR 模板\n```markdown\n# ADR-{序号}: {决策标题}\n\n- 日期: {YYYY-MM-DD}\n- 状态: 提议 / 采纳 / 废弃 / 替代(被ADR-XX替代)\n- 决策者: product-lead + fullstack-dev\n\n## 背景\n{为什么需要做这个决策？当前的问题是什么？}\n\n## 决策\n{我们选择了什么方案？}\n\n## 备选方案\n| 方案 | 优点 | 缺点 | 排除原因 |\n|------|------|------|---------|\n| A | | | |\n| B(采纳) | | | — |\n| C | | | |\n\n## 影响\n- 影响模块: {列出}\n- 迁移成本: 高/中/低/无\n- 风险: {可能的问题}\n- 技术债: {引入了什么债/减少了什么债}\n\n## 后续\n- [ ] {需要做的事}\n```\n\n### ADR 存放位置\n`shared/products/{产品}/adrs/` 目录，或写在 architecture.md 的 ## 决策记录 章节。\n\n### 触发 ADR 的场景\n- 新增/替换框架或主要依赖\n- 数据库 schema 重大变更\n- API 破坏性变更\n- 新增微服务/模块拆分\n- 认证/授权方案变更\n- 部署架构变更\n\n## 代码 Review 清单\n\n### 通用检查（所有PR）\n- [ ] **功能正确**: 满足验收标准，边界情况处理\n- [ ] **无回退**: 不破坏已有功能（跑现有测试）\n- [ ] **命名清晰**: 变量/函数/类名可读、一致\n- [ ] **无硬编码**: 配置项走 env/config，不写死\n- [ ] **错误处理**: try-catch 合理，用户友好错误提示\n- [ ] **日志合理**: 关键操作有日志，无敏感信息泄露\n\n### 安全检查\n- [ ] **输入验证**: 所有用户输入已校验/转义\n- [ ] **SQL注入**: 使用参数化查询/ORM，无拼接SQL\n- [ ] **XSS**: 输出已转义，dangerouslySetInnerHTML 有理由\n- [ ] **CSRF**: 表单有token\n- [ ] **认证/授权**: 路由权限检查，无越权访问\n- [ ] **敏感数据**: 密码哈希、token不明文、日志无敏感信息\n- [ ] **依赖安全**: 无已知漏洞依赖（npm audit / composer audit）\n\n### 性能检查\n- [ ] **N+1查询**: 列表页是否 eager load\n- [ ] **索引**: 新增查询条件是否有对应索引\n- [ ] **大数据**: 分页是否正确，无全表扫描\n- [ ] **缓存**: 高频读取是否命中缓存\n- [ ] **前端**: 组件是否合理 memo/懒加载，bundle size\n\n### 可维护性检查\n- [ ] **单一职责**: 函数/组件不超过 100 行（超了要拆）\n- [ ] **重复代码**: DRY，有无可抽取的公共逻辑\n- [ ] **测试**: 新功能有测试，改动不破坏已有测试\n- [ ] **文档**: 复杂逻辑有注释，API 变更更新文档\n- [ ] **兼容性**: 数据库 migration 可回滚\n\n### Review 严格度（按任务复杂度）\n| 复杂度 | Review 方式 |\n|--------|------------|\n| Simple | 自检通过即可（功能ok + lint pass + tests pass） |\n| Medium | 过通用 + 安全清单 |\n| Complex | 全部清单 + 架构影响评估 + 单独QA |\n\n## 技术债评估框架\n\n### 技术债分类\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **架构债** | 系统结构不合理 | 单体过大、循环依赖、缺中间层 |\n| **代码债** | 代码质量差 | 重复代码、过长函数、magic number |\n| **依赖债** | 过时/不安全依赖 | 大版本落后、弃维库、安全漏洞 |\n| **测试债** | 测试覆盖不足 | 无测试、只有happy path、无E2E |\n| **文档债** | 文档缺失/过时 | API无文档、README过时、部署步骤缺失 |\n| **基础设施债** | DevOps/部署问题 | 手动部署、无CI/CD、无监控 |\n\n### 技术债评分（DORA-lite）\n| 因子 | 说明 | 分值 |\n|------|------|:----:|\n| **影响范围** | 影响多少模块/用户 | 1-5 |\n| **修复难度** | 工时估算 | 1-5（5=最难） |\n| **恶化速度** | 不修会不会越来越糟 | 1-5（5=最快恶化） |\n| **阻塞程度** | 是否阻碍新功能开发 | 1-5（5=严重阻塞） |\n\n**技术债优先级 = 影响范围 × 恶化速度 × 阻塞程度 / 修复难度**\n\n分数越高越优先处理。\n\n### 技术债登记格式\n```markdown\n## TD-{序号}: {标题}\n- 类型: 架构/代码/依赖/测试/文档/基础设施\n- 位置: {文件/模块}\n- 评分: 影响={} 难度={} 恶化={} 阻塞={} → 优先级={}\n- 描述: {具体问题}\n- 修复方案: {怎么修}\n- 预估工时: {}\n- 状态: 登记/计划中/修复中/已完成\n```\n\n### 技术债治理规则\n- **每个sprint分配 20% 时间还债**（product-lead 认可）\n- 新功能不得引入同类型的新债（\"先还再借\"）\n- 每月做一次技术债 review，更新评分和优先级\n- 高恶化速度(≥4)的债立即排期，不等sprint\n\n## Git 工作流\n\n### 分支策略（简化 Git Flow）\n```\nmain          ← 生产分支，只接受 merge\n  └── develop ← 开发主线\n       ├── feature/xxx  ← 功能分支\n       ├── fix/xxx      ← 修复分支\n       └── hotfix/xxx   ← 紧急修复（从main拉，合回main+develop）\n```\n\n### Commit 规范\n```\n<type>(<scope>): <subject>\n\ntype: feat|fix|refactor|docs|style|test|chore|perf\nscope: 模块名（可选）\nsubject: 祈使句，不超过72字符\n```\n\n## 部署清单（上线前）\n- [ ] 所有测试通过\n- [ ] Migration 可回滚\n- [ ] 环境变量已配置\n- [ ] 缓存已清理/预热\n- [ ] 监控告警已设置\n- [ ] 回滚方案已确认\n- [ ] 关键路径手动验证\n\nFile v2.0.6:references/agent-refs/growth-lead/methodology.md\n\n# 增长方法论\n\n## AARRR 海盗指标\n\n| 阶段 | 核心指标 | 当前优先级 |\n|------|---------|:----------:|\n| **A**cquisition 获客 | 新访客、流量来源 | ⭐⭐⭐ |\n| **A**ctivation 激活 | 注册率、首次核心操作完成率 | ⭐⭐ |\n| **R**etention 留存 | D1/D7/D30 留存 | ⭐⭐ |\n| **R**evenue 收入 | ARPU、付费转化 | ⭐（暂缓） |\n| **R**eferral 推荐 | NPS、分享率 | ⭐ |\n\n**当前重心：Acquisition + Activation（冷启动期）**\n\n## ICE 实验评分\n\n| 因子 | 说明 | 范围 |\n|------|------|------|\n| **I**mpact | 成功后对指标影响多大 | 1-10 |\n| **C**onfidence | 证据/经验支撑 | 1-10 |\n| **E**ase | 实施难度（10=最易） | 1-10 |\n\n**ICE = (I+C+E) / 3**\n\n## 实验记录格式\n\n```markdown\n## 实验: {标题}\n- ICE: I=_ C=_ E=_ → 分数: _\n- AARRR环节: {哪个}\n- 假设: 如果{做法}，则{指标}提升{预期}%\n- 结果: {数据} → 成功/失败/待定\n- 教训: {下次改进}\n```\n\n## 零预算策略工具箱\n\n| 策略 | 渠道 | 阶段 | 难度 |\n|------|------|------|------|\n| GEO优化 | AI搜索 | 冷启动 | 中 |\n| 长尾SEO | Google/Bing | 中长期 | 低 |\n| 社区种草 | Reddit/HN/Quora | 冷启动 | 低 |\n| PH Launch | Product Hunt | 里程碑 | 高 |\n| 开源引流 | GitHub | 中期 | 中 |\n| 知识图谱占位 | Wiki/Wikidata/G2 | 早期 | 中 |\n\n## 向 product-lead 提需求的规则\n1. **数据依据**：哪些数据表明需要改动？\n2. **预期影响**：哪个指标提升多少？\n3. **不做的代价**？\n4. **最小验证**：能不改代码先验证吗？\n\n**原则：能不动代码就不动代码。增长优先选内容和运营。**\n\nFile v2.0.6:references/agent-refs/intel-analyst/methodology.md\n\n# 竞品分析方法论\n\n## 竞品情报四维框架\n\n每次扫描竞品，从四个维度收集：\n\n| 维度 | 关注点 | 信息源 |\n|------|--------|--------|\n| **产品动态** | 新功能、下线功能、API变更、技术栈变化 | Changelog、博客、GitHub |\n| **商业动态** | 定价变更、融资、合作、收购、裁员 | 新闻、Crunchbase、LinkedIn |\n| **用户声音** | 好评主题、差评主题、迁移原因、NPS趋势 | G2、Capterra、PH、Reddit |\n| **营销动态** | 渠道策略、内容策略、社媒活跃度、广告投放 | SimilarWeb、社媒、搜索 |\n\n## 威胁/机会评估矩阵\n\n每个重要发现必须评估：\n\n| 评估项 | 选项 |\n|--------|------|\n| **性质** | 🔴 威胁 / 🟢 机会 / 🟡 中性 |\n| **紧迫度** | 立即(本周) / 近期(本月) / 观察(本季) |\n| **影响度** | 高(影响核心业务) / 中(影响部分功能) / 低(边缘) |\n| **建议行动** | 防御(我们也要做) / 差异化(做他们不做的) / 忽略 |\n\n## 竞品档案模板\n\n```markdown\n## {竞品名}\n- 定位: {一句话}\n- 主要用户群: {谁在用}\n- 定价: {免费/付费模式}\n- 核心优势: {为什么用户选它}\n- 核心劣势: {用户最大抱怨}\n- 威胁等级: 🔴高 / 🟡中 / 🟢低\n- 最近动态: {最新1-2条}\n- 我们的差异化: {我们做了什么它没做的}\n```\n\n## 扫描优先级\n- **周一全量扫描**: 所有竞品四维全扫\n- **周三/周五快扫**: Top 3 威胁竞品，只看产品+商业动态\n- **>5个竞品时**: Top 3 深扫，其余浅扫（只查新闻）\n\nArchive v2.0.4: 18 files, 55626 bytes\n\nFiles: README.md (4997b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (14240b), references/soul-templates.md (19800b), scripts/deploy.js (21405b), SKILL.md (25541b), team-builder.json (394b), _meta.json (131b)\n\nFile v2.0.4:SKILL.md\n\n---\r\nname: team-builder\r\ndescription: Deploy a multi-agent SaaS growth team on OpenClaw with shared workspace, real-time spawn dispatch (primary) plus async inbox (backup), cron-scheduled tasks, deep project code scanning (Deep Dive), and optional Telegram integration. Use when building or upgrading multi-agent teams for SaaS/product-matrix work. Supports dual-development tracks by default: `devops` for delivery/deploy/environment/acceptance and `fullstack-dev` for implementation/module deep-dive/claude-only coding execution using direct acpx or existing session continuity. Includes Project Deep Dive capability so shared product knowledge files (DB schema, routes, models, services, auth, integrations, tech debt, etc.) can be generated and consumed efficiently by all agents. Supports customizable team name, agent roles, models, timezone, and Telegram bots.\r\n---\r\n\r\n# Team Builder\r\n\r\nDeploy a reusable multi-agent SaaS/growth team template on OpenClaw in one shot.\r\n\r\n## System Impact & Prerequisites\r\n\r\n> **Read before running.** This skill creates files and modifies system config.\r\n\r\n### What it creates\r\n- A new workspace directory with ~40 files (agent configs, shared knowledge, inboxes, kanban)\r\n- `apply-config.js` -- script that **modifies `~/.openclaw/openclaw.json`** (adds agents, bindings, agentToAgent config). Auto-backs up before writing.\r\n- `create-crons.ps1` / `create-crons.sh` -- scripts that **create cron jobs** via `openclaw cron add`\r\n- After running these scripts you must **restart the gateway** (`openclaw gateway restart`)\r\n\r\n### What it does NOT do automatically\r\n- Does not modify openclaw.json directly -- you run `apply-config.js` yourself\r\n- Does not create cron jobs directly -- you run the cron script yourself\r\n- Does not restart the gateway -- you do that manually\r\n\r\n### Optional: Telegram\r\n- If you provide bot tokens during setup, `apply-config.js` will also add Telegram account configs and bindings\r\n- Requires: Telegram bot tokens from @BotFather, your Telegram user ID\r\n- Requires: network access to Telegram API (proxy configurable)\r\n\r\n### Optional: ACP / Claude Code\r\n- The `fullstack-dev` agent is configured as the implementation-focused Claude coding role\r\n- Current production path is **claude only**\r\n- Preferred execution modes: simple direct, medium Claude ACP `run` or direct acpx, complex work via existing fullstack-dev continuity + context files\r\n- Do not assume IM-bound ACP `session` persistence is available\r\n\r\n### Internal Dispatch Protocol (MANDATORY)\r\n- **Standard agents** (product-lead, growth-lead, intel-analyst, devops, etc.): dispatch via `sessions_spawn(runtime=\"subagent\", mode=\"run\")` — **never include `streamTo` parameter**\r\n- **ACP agents** (fullstack-dev): dispatch via `sessions_spawn(runtime=\"acp\")` — may include `streamTo=\"parent\"`\r\n- Results arrive via auto-announce push; do not poll with `sessions_list` or `subagents list`\r\n- Chief-of-staff is the single group entry point; other agents are spawned internally, not expected to independently monitor group chat\r\n- See `shared/knowledge/team-workflow.md` 零章 for the canonical dispatch specification\r\n\r\n### Credentials involved\r\n- **Telegram bot tokens** (optional) -- stored in openclaw.json, used for agent-to-Telegram binding\r\n- **Model API keys** -- must already be configured in your OpenClaw model providers (not handled by this skill)\r\n\r\n### Recommended\r\n- Review generated `apply-config.js` before running\r\n- Check the backup of openclaw.json after running\r\n- Test with 2-3 agents before enabling all cron jobs\r\n\r\n## Team Architecture\r\n\r\nDefault reference architecture for a SaaS/growth multi-agent team (customizable to 2-10 agents):\r\n\r\n```\r\nCEO\r\n |-- Chief of Staff (dispatch + strategy + efficiency)\r\n |-- Data Analyst (data + user research)\r\n |-- Growth Lead (GEO + SEO + community + social media)\r\n |-- Content Chief (strategy + writing + copywriting + i18n)\r\n |-- Intel Analyst (competitor monitoring + market trends)\r\n |-- Product Lead (product management + tech architecture)\r\n |-- DevOps (delivery / deploy / environment / acceptance)\r\n |-- Fullstack Dev (implementation / module deep dive / ACP coding session)\r\n```\r\n\r\n### Multi-Team Support\r\n\r\nOne OpenClaw instance can run multiple teams:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js                  # default team\r\nnode <skill-dir>/scripts/deploy.js --team alpha      # named team \"alpha\"\r\nnode <skill-dir>/scripts/deploy.js --team beta       # named team \"beta\"\r\n```\r\n\r\nNamed teams use prefixed agent IDs (`alpha-chief-of-staff`, `beta-growth-lead`) to avoid conflicts. Each team gets its own workspace subdirectory.\r\n\r\n### Flexible Team Size\r\n\r\nThe wizard lets you select 2-10 agents from the available roles. Skip roles you don't need. The 8-agent default covers most SaaS scenarios with dual-dev routing, but you can run leaner (3-4 agents) or expand with custom roles.\r\n\r\n### Model Auto-Detection\r\n\r\nThe wizard scans your `openclaw.json` for registered model providers and auto-suggests models by role type:\r\n\r\n| Role Type | Best For | Auto-detect Pattern |\r\n|-----------|----------|-------------------|\r\n| Thinking | Strategic roles (chief, growth, content, product) | /glm-5\\|opus\\|o1\\|deepthink/i |\r\n| Execution | Operational roles (data, intel, fullstack) | /glm-4\\|sonnet\\|gpt-4/i |\r\n| Fast | Lightweight tasks | /flash\\|haiku\\|mini/i |\r\n\r\nYou can always override with manual model IDs.\r\n\r\n## Setup / Config / Scripts\r\n\r\n### Required Inputs\r\n- Team name\r\n- Workspace dir\r\n- Timezone\r\n- Morning brief hour\r\n- Evening brief hour\r\n- Thinking model\r\n- Execution model\r\n- CEO title\r\n\r\n### Optional Inputs\r\n- Telegram user ID\r\n- Telegram bot tokens\r\n- Proxy\r\n- ACP coding agent（给 fullstack-dev 使用）\r\n\r\n### Core Scripts\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js\r\nnode <workspace-dir>/apply-config.js\r\npowershell <workspace-dir>/create-crons.ps1\r\nbash <workspace-dir>/create-crons.sh\r\nopenclaw gateway restart\r\n```\r\n\r\n### Execution Priority\r\n- First: matched execution skill (for coding work, `coding-lead` if loaded)\r\n- Second: agent-role fallback when no matching skill is loaded\r\n- Third: templates/README explain boundaries and ownership only; they should not override matched skills\r\n\r\n### Context File Hygiene\r\n- Active context files live under `<project>/.openclaw/`\r\n- Reuse one context file per active code chain when possible\r\n- Naming pattern: `context-<task-slug>.md`\r\n- Active context file cap per project: **60**\r\n- Context-file lifecycle window per project: **100 total files** across active + archive\r\n- Completed or stale files should be deleted or moved to `.openclaw/archive/`\r\n\r\n### Current Dual-Dev Standard\r\n- fullstack-dev：实现、模块深挖、开发文档、接口文档、Claude coding 执行；默认 coding skill 可采用 `coding-lead`，其中 simple 任务直做，medium 倾向 Claude ACP `run` 或 direct acpx，complex 通过现有会话连续协作 + 上下文文件推进，不把 ACP `session` 持久线程作为正式主路径；context 活跃上限 60、生命周期总窗口 100；并行允许但必须先定义边界，总上限 5 个工作单元\r\n- devops：交付、部署、环境、回归、冒烟、自动QA、发布门禁\r\n- product-lead：澄清、PRD、验收标准，不完整不得派工\r\n- chief-of-staff：路由、裁决、控制 token 浪费\r\n\r\n## Deployment Flow\r\n\r\n### Step 1: Collect Configuration\r\n\r\nAsk the user for these inputs (use defaults if not provided):\r\n\r\n| Parameter | Default | Description |\r\n|-----------|---------|-------------|\r\n| Team name | Alpha Team | Used in all docs and configs |\r\n| Workspace dir | `~/.openclaw/workspace-team` | Shared workspace root |\r\n| Timezone | Asia/Shanghai | For cron schedules |\r\n| Morning brief hour | 8 | Chief's morning report |\r\n| Evening brief hour | 18 | Chief's evening report |\r\n| Thinking model | zai/glm-5 | For strategic roles |\r\n| Execution model | zai/glm-4.7 | For execution roles |\r\n| CEO title | Boss | How agents address the CEO |\r\n\r\nOptional: Telegram user ID, proxy, and 7 bot tokens.\r\n\r\n### Step 2: Run Deploy Script\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js\r\n```\r\n\r\nInteractive -- asks all questions from Step 1, generates the full workspace.\r\n\r\n### Step 2b: Non-interactive / Verify Mode\r\n\r\nPrepare a JSON config file:\r\n\r\n```json\r\n{\r\n  \"teamName\": \"Alpha Team\",\r\n  \"workspaceDir\": \"~/.openclaw/workspace-team\",\r\n  \"timezone\": \"Asia/Shanghai\",\r\n  \"morningHour\": 8,\r\n  \"eveningHour\": 18,\r\n  \"thinkingModel\": \"zai/glm-5\",\r\n  \"executionModel\": \"zai/glm-4.7\",\r\n  \"ceoTitle\": \"Boss\",\r\n  \"roles\": [\"chief-of-staff\",\"data-analyst\",\"growth-lead\",\"content-chief\",\"intel-analyst\",\"product-lead\",\"devops\",\"fullstack-dev\"]\r\n}\r\n```\r\n\r\nRun:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js --config team-builder.json\r\nnode <skill-dir>/scripts/deploy.js --verify --config team-builder.json\r\n```\r\n\r\n`--verify` checks that generated files contain the expected dual-dev model, role ownership, and cron entries.\r\n\r\n### Step 3: Apply Config\r\n\r\n```bash\r\nnode <workspace-dir>/apply-config.js\r\n```\r\n\r\nAdds agents to openclaw.json, preserving existing config.\r\n\r\n### Step 4: Create Cron Jobs\r\n\r\n```bash\r\n# Windows\r\npowershell <workspace-dir>/create-crons.ps1\r\n\r\n# Linux/Mac\r\nbash <workspace-dir>/create-crons.sh\r\n```\r\n\r\n### Step 5: Restart Gateway\r\n\r\n```bash\r\nopenclaw gateway restart\r\n```\r\n\r\n### Step 6: Fill Business Info\r\n\r\nUser must edit:\r\n- `shared/decisions/active.md` -- strategy, priorities\r\n- `shared/products/_index.md` -- products overview (≤5 lines per product: URL, code path, positioning, tech, status). Detailed info goes in each product's `overview.md`.\r\n- `shared/knowledge/competitor-map.md` -- competitor analysis\r\n- `shared/knowledge/tech-standards.md` -- coding standards\r\n\r\n### Step 7: Trigger Deep Dive Scans\r\n\r\nAfter filling in products with code directories, tell product-lead to trigger Deep Dive scans:\r\n1. Product-lead dispatches delivery-oriented scan requests to devops (via `sessions_spawn` if online, inbox as fallback)\r\n2. Devops enters each project directory and generates shared knowledge / deployment-oriented scan outputs\r\n3. Fullstack-dev picks up module-level deep dive or implementation follow-up when needed\r\n4. Product-lead reviews the generated files for completeness and acceptance impact\r\n5. All agents now have deep project understanding for informed decisions\r\n\r\n## Cron Schedule\r\n\r\n| Offset | Agent | Task | Frequency |\r\n|--------|-------|------|-----------|\r\n| H-1 | Data Analyst | Data + user feedback | Daily |\r\n| H-1 | Intel Analyst | Competitor scan | Mon/Wed/Fri |\r\n| H | Chief of Staff | Morning brief (announced) | Daily |\r\n| H+1 | Growth Lead | GEO + SEO + community | Daily |\r\n| H+1 | Content Chief | Weekly content plan | Monday |\r\n| H+2 | DevOps | Delivery / environment / Deep Dive / acceptance | Daily |\r\n| H+10 | Chief of Staff | Evening brief (announced) | Daily |\r\n\r\n(H = morning brief hour)\r\n\r\n## Generated File Structure\r\n\r\n```\r\n<workspace>/\r\n├── AGENTS.md, SOUL.md, USER.md  (auto-injected)\r\n├── apply-config.js, create-crons.ps1/.sh, README.md\r\n├── agents/<8 agent dirs>/       (SOUL.md + MEMORY.md + memory/)\r\n└── shared/\r\n    ├── briefings/, decisions/, inbox/ (v2: with status tracking)\r\n    ├── status/team-dashboard.md     (chief-of-staff maintains, all agents read first)\r\n    ├── data/                        (public data pool, data-analyst writes, all read)\r\n    ├── kanban/, knowledge/\r\n    └── products/\r\n        ├── _index.md                (product matrix overview)\r\n        ├── _template/               (knowledge directory template)\r\n        └── {product}/               (per-product knowledge, up to 20 files)\r\n            ├── overview.md, architecture.md, database.md, api.md, routes.md\r\n            ├── models.md, services.md, frontend.md, auth.md, integrations.md\r\n            ├── jobs-events.md, config-env.md, dependencies.md, devops.md\r\n            ├── test-coverage.md, tech-debt.md, domain-flows.md, data-flow.md\r\n            ├── i18n.md, changelog.md, notes.md\r\n```\r\n\r\n\r\n\r\n## Knowledge Governance\r\n\r\nEach shared knowledge file has a designated owner. Only the owner agent updates it; others read only.\r\n\r\n| File | Owner | Update Trigger |\r\n|------|-------|---------------|\r\n| geo-playbook.md | growth-lead | After GEO experiments/discoveries |\r\n| seo-playbook.md | growth-lead | After SEO experiments |\r\n| competitor-map.md | intel-analyst | After each competitor scan |\r\n| content-guidelines.md | content-chief | After proven writing patterns |\r\n| user-personas.md | data-analyst | After new user insights |\r\n| tech-standards.md | product-lead | After architecture decisions |\r\n\r\n### Update Protocol\r\nWhen updating a knowledge file, the owner must:\r\n1. Add a dated entry at the top: `## [YYYY-MM-DD] <what changed>`\r\n2. Include the reason and data evidence\r\n3. Never delete existing entries without CEO approval (append, don't replace)\r\n\r\n### Chief of Staff Governance\r\nThe chief-of-staff monitors knowledge file health during weekly reviews:\r\n- Are files being updated regularly?\r\n- Any conflicting information between files?\r\n- Any stale entries that should be archived?\r\n\r\n## Self-Evolution Pattern\r\n\r\nAgents improve their own strategies over time through a feedback loop:\r\n\r\n```\r\n1. Execute task (cron or inbox triggered)\r\n2. Collect results (data, metrics, outcomes)\r\n3. Analyze: what worked vs what didn't\r\n4. Update knowledge files with proven strategies (with evidence)\r\n5. Next execution reads updated knowledge → better performance\r\n```\r\n\r\nThis is NOT the agent randomly changing rules. Updates must be:\r\n- **Data-driven**: backed by metrics or concrete outcomes\r\n- **Incremental**: append new findings, don't rewrite everything\r\n- **Traceable**: dated with evidence so others can verify\r\n\r\n### What Agents Can Self-Update\r\n- Their own knowledge files (per ownership table above)\r\n- Their own MEMORY.md (lessons learned, decisions)\r\n- shared/data/ outputs (data-analyst only)\r\n\r\n### What Requires CEO Approval\r\n- shared/decisions/active.md (strategy changes)\r\n- Adding/removing agents or changing team architecture\r\n- External publishing or spending decisions\r\n\r\n## Public Data Layer\r\n\r\nThe `shared/data/` directory serves as a read-only data pool for all agents:\r\n\r\n- **data-analyst** writes: daily metrics, user feedback summaries, anomaly alerts\r\n- **All agents** read: to inform their own decisions\r\n- Format: structured markdown or JSON, dated filenames (e.g., `metrics-2026-03-01.md`)\r\n- Retention: keep 30 days, archive older files\r\n\r\n## Project Deep Dive — Code Scanning\r\n\r\nAgents can deeply understand each SaaS product through automated code scanning. This is critical — without deep project knowledge, all team decisions are surface-level.\r\n\r\n### How It Works\r\n\r\n1. CEO adds a product to `shared/products/_index.md` (name, URL, code directory, tech stack)\r\n2. Product Lead triggers a delivery-oriented Deep Dive scan by dispatching to DevOps (via `sessions_spawn` if online, inbox as fallback)\r\n3. DevOps enters the project directory (read-only) and generates shared knowledge / delivery-oriented scan outputs\r\n4. Fullstack Dev picks up module-level deep dive or implementation follow-up when needed\r\n5. Knowledge files are generated in `shared/products/{product}/`\r\n6. All agents consume these files **via manifest-based lazy loading** (never read all at once)\r\n\r\n### Manifest-Based Lazy Loading (MANDATORY)\r\n\r\nEach product directory includes a `manifest.json` (~200 tokens) that lists all files with one-line summaries and a `taskFileMap` mapping task types to relevant files.\r\n\r\n**Agent workflow:**\r\n1. Read `_index.md` → identify which product\r\n2. Read `{product}/manifest.json` → see all files + summaries (~200 tokens)\r\n3. Based on `taskFileMap` or summaries, read only 1-3 relevant files\r\n4. Never read more than 5 product files per session\r\n\r\n**Why:** With 15+ products × 20 files each, full loading = 40K+ tokens per product. Manifest loading = 200 tokens + only what's needed.\r\n\r\n**DevOps MUST regenerate `manifest.json`** after every delivery-oriented scan (L0-L4). Fullstack Dev updates it when doing module-level follow-up that changes knowledge scope. Template in `_template/manifest.json`.\r\n\r\n### Manifest Quality Standards\r\n\r\n摘要不能为了省 token 丢掉关键信息。每条摘要须满足：\r\n- **核心文件**（database/models/services/routes/integrations）：50-130字，列出关键实体名/数量/域名\r\n- **中等文件**（auth/frontend/commands/config）：30-80字，点明方案和范围\r\n- **轻量文件**（changelog/notes/metrics）：可以短（<20字）\r\n- **taskFileMap**：必须覆盖该产品的所有核心业务场景（不少于8个映射）\r\n- **codeStats**：必须包含文件数、行数、模型数、表数等量化指标\r\n\r\n### Product Knowledge Directory\r\n\r\nEach product gets a knowledge directory with up to 20 files + manifest:\r\n\r\n```\r\nshared/products/{product}/\r\n├── manifest.json        ← **INDEX** (~200 tokens): file list, summaries, taskFileMap\r\n├── overview.md          ← Product positioning (from _index.md)\r\n├── architecture.md      ← System architecture, tech stack, design patterns, layering\r\n├── database.md          ← Full table schema, relationships, indexes, migrations\r\n├── api.md               ← API endpoints, params, auth, versioning\r\n├── routes.md            ← Complete route table (Web + API + Console)\r\n├── models.md            ← ORM relationships, scopes, accessors, observers\r\n├── services.md          ← Business logic, state machines, workflows, validation\r\n├── frontend.md          ← Component tree, page routing, state management\r\n├── auth.md              ← Auth scheme, roles/permissions matrix, OAuth\r\n├── integrations.md      ← Third-party: payment/email/SMS/storage/CDN/analytics\r\n├── jobs-events.md       ← Queue jobs, event listeners, scheduled tasks, notifications\r\n├── config-env.md        ← Environment variables, feature flags, cache strategy\r\n├── dependencies.md      ← Key dependencies, custom packages, vulnerabilities\r\n├── devops.md            ← Deployment, CI/CD, Docker, monitoring, logging\r\n├── test-coverage.md     ← Test strategy, coverage, weak spots\r\n├── tech-debt.md         ← TODO/FIXME/HACK inventory, dead code, complexity hotspots\r\n├── domain-flows.md      ← Core user journeys, domain boundaries, module coupling\r\n├── data-flow.md         ← Data lifecycle: external → import → process → store → output\r\n├── i18n.md              ← Internationalization, language coverage\r\n├── changelog.md         ← Scan diff log (what changed between scans)\r\n└── notes.md             ← Agent discoveries, gotchas, implicit rules\r\n```\r\n\r\n### Scan Levels\r\n\r\n| Level | Scope | When | Output |\r\n|-------|-------|------|--------|\r\n| L0 Snapshot | Surface: directory tree, packages, env | First onboard | architecture, dependencies, config-env |\r\n| L1 Skeleton | Structure: DB, routes, models, components | First onboard | database, routes, api, models, frontend |\r\n| L2 Deep Dive | Logic: services, auth, jobs, integrations | On-demand per module | services, auth, jobs-events, integrations, domain-flows, data-flow |\r\n| L3 Health Check | Quality: tech debt, tests, security | Periodic / pre-release | tech-debt, test-coverage, devops |\r\n| L4 Incremental | Delta: git diff → update affected files | After code changes | changelog + targeted updates |\r\n\r\n### Content Standards\r\n\r\nKnowledge files capture not just WHAT exists but WHY:\r\n- **Design decisions**: Why this approach was chosen\r\n- **Implicit business rules**: Logic buried in code (e.g., \"orders auto-cancel after 72h\")\r\n- **Gotchas**: What breaks if you touch this module carelessly\r\n- **Cross-module coupling**: Where changing A silently breaks B\r\n- **Performance hotspots**: N+1 queries, missing indexes, bottleneck endpoints\r\n\r\n### Role Responsibilities\r\n\r\n| Role | Responsibility |\r\n|------|---------------|\r\n| Product Lead | **Clarification / PRD / acceptance**: complete clarification, PRD, user stories, acceptance criteria, and review knowledge freshness before delegating |\r\n| DevOps | **Delivery / QA gate / Deep Dive**: enter code directory for deployment-oriented scans, maintain release checklist, smoke/regression testing, auto-QA access, and generate/update shared product knowledge files |\r\n| Fullstack Dev | **Implementation / docs / Deep Dive follow-up**: continue module-level deep dive, code analysis, implementation, dev docs, interface docs, and ACP session work |\r\n| Chief of Staff | **Routing / escalation**: split implementation vs delivery tasks, prevent duplicate labor, escalate blockers |\r\n| All Agents | **Consumption**: read product knowledge before any product-related decision |\r\n\r\n### Per-Stack Auto-Detection\r\n\r\nFullstack Dev auto-detects tech stack and applies stack-specific scan strategies:\r\n- **Laravel/PHP**: migrations, route:list, Models, Services, Middleware, Policies, Jobs, Console/Kernel\r\n- **React/Vue**: components, router, stores, API client, i18n\r\n- **Python/Django/FastAPI**: models.py, urls.py, views.py, middleware, celery\r\n- **General**: tree, git log, grep TODO/FIXME, .env.example, Docker, CI, tests\r\n\r\n## Team Coordination v2\r\n\r\n### Inbox Protocol v2 (backup channel, status tracking)\r\n\r\n> **Primary dispatch**: `sessions_spawn` (real-time). Inbox is for archival, cross-session handoff, and fallback when spawn is unavailable.\r\n\r\nEvery inbox message now has a `status` field:\r\n- `pending` → `received` → `in-progress` → `done` (or `blocked`)\r\n- Chief-of-staff monitors timeouts: high>4h, normal>24h pending = intervention\r\n- Blocked >8h = escalation to CEO\r\n- Recipients MUST update status immediately upon reading\r\n\r\n### Team Dashboard (`shared/status/team-dashboard.md`)\r\n\r\nChief-of-staff maintains a \"live scoreboard\" updated every session:\r\n- 🔴 Urgent/Blocked items\r\n- 📊 Per-agent status table (last active, current task, status icon)\r\n- 📬 Unprocessed inbox summary (pending/blocked messages across all inboxes)\r\n- 🔗 Cross-agent task chain tracking (A→B→C with per-step status)\r\n- 📅 Today/Tomorrow focus\r\n\r\n**All agents read this file first when waking up.** 5-second situational awareness.\r\n\r\n### Chief-of-Staff as Router\r\n\r\nThe chief is upgraded from \"briefing writer\" to \"active team router\":\r\n- **Real-time dispatch**: uses `sessions_spawn(runtime=\"subagent\")` to directly wake agents and assign tasks — this is the primary dispatch method\r\n- **Blocker detection**: scans all inboxes for overdue messages\r\n- **Inbox as backup**: writes to inbox only for archival, cross-session handoff, or when agent is unreachable\r\n- **Task chain tracking**: identifies multi-agent workflows and tracks each step\r\n- **Escalation**: persistent blockers get flagged to CEO\r\n- **Runs 4x/day** (morning brief, midday patrol, afternoon patrol, evening brief)\r\n\r\n### Cron Schedule (10 jobs, up from 7)\r\n\r\n| Time | Agent | Type | Purpose |\r\n|------|-------|------|---------|\r\n| 07:00 | data-analyst | daily | Data pull + feedback scan |\r\n| 08:00 | chief-of-staff | **announce** | Morning: router scan + brief + quality |\r\n| 09:00 | growth-lead | daily | GEO/SEO/community |\r\n| 09:00 | product-lead | **daily (NEW)** | Inbox + clarification/PRD + task delegation |\r\n| 10:00 | content-chief | **daily M-F (was weekly)** | Content creation + collaboration |\r\n| 10:00 | devops | **daily (delivery track)** | Inbox + Deep Dive + delivery + QA gate |\r\n| 12:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 15:00 | chief-of-staff | **patrol (NEW)** | Router scan only, no brief |\r\n| 18:00 | chief-of-staff | **announce** | Evening: router scan + summary + next day plan |\r\n| 07:00 M/W/F | intel-analyst | 3x/week | Competitor scan |\r\n\r\n### Why These Changes Matter\r\n\r\n| Before | After | Impact |\r\n|--------|-------|--------|\r\n| Inbox = primary dispatch | Inbox = backup + spawn = primary | Real-time dispatch via spawn; inbox for archival only |\r\n| Chief 2x/day | Chief 4x/day with router role | Blockers caught within hours, not days |\r\n| Content-chief 1x/week | Daily M-F | Actually produces content |\r\n| Product-lead no cron | Daily | Knowledge governance happens |\r\n| No team dashboard | Dashboard every session | All agents know the full picture |\r\n| No timeout detection | Automatic timeout rules | Nothing falls through cracks |\r\n\r\n## Key Design Decisions\r\n\r\n- **Shared workspace** so qmd indexes everything for all agents\r\n- **Real-time spawn dispatch** as primary inter-agent communication; inbox as backup for archival and cross-session handoff\r\n- **Chief as Router** — active coordinator who dispatches via `sessions_spawn`, detects blockers, and resolves them\r\n- **Team Dashboard** — single source of truth for team-wide status, maintained by chief every session\r\n- **GEO as #1 priority** (AI search = blue ocean)\r\n- **Fullstack Dev spawns Claude Code** via ACP for complex implementation tasks\r\n- **DevOps owns delivery and QA gate** so implementation and release responsibilities stay separated\r\n- **Project Deep Dive** gives all agents deep codebase understanding, not just surface-level product overviews\r\n\r\n## Customization\r\n\r\nEdit ROLES array in `scripts/deploy.js` to add/remove agents.\r\nEdit `references/soul-templates.md` for SOUL.md templates.\r\nEdit `references/shared-templates.md` for shared file templates.\n\nFile v2.0.4:README.md\n\n# team-builder\n\n## 这是什么\n`team-builder` 用于在 OpenClaw 上生成一套面向 SaaS / 产品矩阵场景的多 Agent 团队工作区与配套脚本。\n\n它负责输出:\n- 团队工作区目录\n- Agent 基础文件\n- shared 协作文件\n- `apply-config.js`\n- `create-crons.ps1` / `create-crons.sh`\n- 角色与协作模板\n\n当前默认采用一套**可复用的双开发轨参考架构**:\n- `devops`:交付、部署、环境、验收、交付向 Deep Dive\n- `fullstack-dev`:项目内实现、模块深挖、代码链连续开发\n\n注意:这是一类多 agent 团队模板,不要求你的团队必须由 team-builder 创建后才成立。\n\n## 适用范围\n适用于:\n- 新建 OpenClaw 团队工作区\n- 升级现有 SaaS 多 Agent 团队模板\n- 为已经存在的多 agent 团队抽取或对齐一套标准化模板\n- 需要内置信箱协作、dashboard、产品知识目录、Deep Dive 流程的团队\n\n不适用于:\n- 直接替你修改线上环境\n- 自动执行 gateway 重启、cron 创建、openclaw.json 写入(这些由你确认后手动执行生成脚本)\n\n## 核心能力\n\n它识别和生成的是\"此类多 agent 协作团队模板\",不是\"唯一正确的团队来源\"。如果你已经有多 agent 团队,只要结构接近,也可以直接吸收它的工作流规则。\n\n\n### 1. 一次生成团队骨架\n生成共享工作区与角色目录,包括:\n- `AGENTS.md`\n- `SOUL.md`\n- `USER.md`\n- `shared/`\n- `agents/`\n- 配置 / cron 脚本\n\n### 2. 双开发轨模板\n内置:\n- `product-lead` 负责澄清 / PRD / 验收 / 路由\n- `devops` 负责交付 / QA gate / delivery-oriented Deep Dive\n- `fullstack-dev` 负责实现 / 模块级 deep dive follow-up / 编码执行\n\n### 3. Deep Dive 产品知识目录\n生成适合沉淀产品知识的共享目录结构,并支持 manifest 驱动的延迟阅读模式。\n\n### 4. 非交互 + verify\n支持:\n- 交互式部署\n- `--config` 非交互部署\n- `--verify` 校验生成结果\n\n当前 `--verify` 会输出**结构化 JSON 检查报告**,覆盖:\n- core\n- roles\n- fallback\n- workflow\n- hygiene\n- cron\n\n### 5. fallback 与上下文治理同步\n已对齐当前编码执行口径:\n- Claude-only ACP 默认口径\n- context active cap:**60**\n- lifecycle window:**100**\n- done 前 verify\n- 写入 / spawn 前确认 `cwd`\n\n## 初始化\n本技能**无需单独初始化**,但第一次使用前建议确认:\n- OpenClaw 已正确安装\n- Node.js 可用\n- 你有权查看并手动执行生成出来的脚本\n\n## 使用方法\n\n### 1. 交互式生成\n```bash\nnode <skill-dir>/scripts/deploy.js\n```\n适合首次搭团队。\n\n### 2. 非交互生成\n准备一个 JSON 配置文件,再执行:\n```bash\nnode <skill-dir>/scripts/deploy.js --config team-builder.json\n```\n\n### 3. 验证生成结果\n```bash\nnode <skill-dir>/scripts/deploy.js --verify --config team-builder.json\n```\n输出结构化 JSON 检查报告;若失败,进程以非 0 退出。\n\n### 4. 应用生成结果\n生成完成后,按顺序手动执行:\n1. `node <workspace-dir>/apply-config.js`\n2. `powershell <workspace-dir>/create-crons.ps1` 或 `bash <workspace-dir>/create-crons.sh`\n3. `openclaw gateway restart`\n\n## 配置说明\n\n### 必填项\n- teamName\n- workspaceDir\n- timezone\n- morningHour\n- eveningHour\n- thinkingModel\n- executionModel\n- ceoTitle\n\n### 可选项\n- roles\n- roleNames\n- `--team` 前缀\n- Telegram 相关输入(如果你准备接入 Telegram)\n\n## 输出物说明\n\n### 1. 工作区\n输出团队工作区,包含:\n- 角色目录\n- inbox / briefings / dashboard\n- products / knowledge / kanban\n- 团队工作流默认规则(任务入口、done 链、最小读取顺序、角色记忆边界)\n\n### 2. `apply-config.js`\n用于把生成的 Agent 写入 `openclaw.json`。\n**注意:不是自动执行,而是生成后由你手动执行。**\n\n### 3. `create-crons.*`\n用于生成 cron 任务。\n**注意:不是自动执行,而是生成后由你手动执行。**\n\n### 4. references/fallback 文件\n生成的团队工作区内现在会包含:\n- `references/coding-behavior-fallback.md`\n\n供 `fullstack-dev` 在 `coding-lead` 未加载时使用。\n\n## 验证建议\n改动本技能时,至少确认:\n- `deploy.js --verify` 仍输出结构化 JSON 报告\n- 生成物里仍包含 fallback 文件和 team-workflow.md\n- README / SKILL / templates / deploy.js 文案一致\n- context 规则仍是 **60 / 100**\n\n## 相关文件\n- `SKILL.md`\n- `scripts/deploy.js`\n- `references/shared-templates.md`\n- `references/soul-templates.md`\n- `references/coding-behavior-fallback.md`\n\n## 最近一次修改(中文)\n- **2026-03-29 / 7502460**\n- 变更摘要:\n  - `deploy.js --verify` 扩展为**结构化 JSON 检查报告**,报告按 `core / roles / fallback / workflow / hygiene / cron` 分组\n  - 生成物自动带 `shared/knowledge/team-workflow.md`(任务入口、done 链、最小读取顺序、角色记忆边界)\n  - `shared-templates.md` 改为指向独立 workflow 规范文件,不再内嵌细则\n  - 定位从\"部署固定 8-agent 团队\"改为\"生成可复用的此类多 agent 团队模板\"\n\nFile v2.0.4:_meta.json\n\n{\n  \"ownerId\": \"kn761rne5jajzfg0d8h6nge4hh8220cq\",\n  \"slug\": \"team-builder\",\n  \"version\": \"2.0.4\",\n  \"publishedAt\": 1774810793219\n}\n\nFile v2.0.4:references/agent-refs/chief-of-staff/dashboard-template.md\n\n# 团队面板模板\n\n```markdown\n# 团队状态面板\n> 参谋长维护 | 最后更新: YYYY-MM-DD HH:MM\n\n## 🔴 紧急/阻塞\n- [描述] — 涉及: [agent] — 原因: [原因] — 已持续: [时长]\n\n## 📊 各agent状态\n| Agent | 上次活跃 | 当前任务 | 状态 |\n|-------|---------|---------|------|\n| 参谋长 | 刚才 | 面板+简报 | ✅ |\n| 产品总监 | | | |\n| 增长总监 | | | |\n| 内容主编 | | | |\n| 数据分析 | | | |\n| 情报分析 | | | |\n| 全栈工程 | | | |\n\n状态: ✅正常 🔄进行中 🔴阻塞 ⏳等待 💤失联(>48h)\n\n## 📬 未处理信箱\n| 消息 | 发→收 | 优先级 | 已等待 | 状态 |\n\n## 🔗 跨agent任务链\n- [任务]: A(✅) → B(🔄) → C(⏳)\n\n## 📅 今日/明日重点\n1. ...\n```\n\nFile v2.0.4:references/agent-refs/chief-of-staff/strategy-methodology.md\n\n# 矩阵策略方法论\n\n## 产品组合评估框架\n\n### 产品健康度评分卡（每产品，月度更新）\n\n| 维度 | 指标 | 权重 | 评分(1-5) | 说明 |\n|------|------|:----:|:---------:|------|\n| **流量** | 月UV增长率 | 20% | | >20%=5, 10-20%=4, 0-10%=3, 持平=2, 下降=1 |\n| **留存** | D7回访率 | 20% | | >40%=5, 30-40%=4, 20-30%=3, 10-20%=2, <10%=1 |\n| **变现** | 收入/利润趋势 | 15% | | 盈利=5, 接近盈亏=4, 有收入=3, 无收入=2, 亏损=1 |\n| **竞争** | 差异化强度 | 15% | | 独占优势=5, 明显差异=4, 有差异=3, 同质=2, 落后=1 |\n| **效率** | 投入产出比 | 15% | | 高ROI=5, 正向=4, 中等=3, 低=2, 负=1 |\n| **战略** | 对矩阵协同价值 | 15% | | 核心枢纽=5, 高协同=4, 有协同=3, 独立=2, 拖累=1 |\n\n**总分 = Σ(评分 × 权重)**\n\n### 产品分类（简化BCG矩阵）\n\n| 类型 | 条件 | 策略 |\n|------|------|------|\n| ⭐ **明星** | 高增长 + 高份额（或潜力） | 全力投入，抢市场 |\n| 🐄 **现金牛** | 低增长 + 高份额 | 维护为主，收割利润，少投入 |\n| ❓ **问号** | 高增长 + 低份额 | 限期验证：3个月达标→升星，否则降级 |\n| 🐕 **瘦狗** | 低增长 + 低份额 | 最低维护或考虑下线，资源释放 |\n\n### 产品组合仪表盘模板\n```markdown\n# 产品矩阵健康度 — {YYYY}-{MM}\n\n| 产品 | 类型 | 健康度 | 本月重点 | 资源分配 |\n|------|------|:------:|---------|:--------:|\n| Richlyva | ❓问号 | 3.2/5 | 冷启动验证 | 40% |\n| ... | | | | |\n\n## 矩阵级风险\n- {跨产品风险}\n\n## 协同机会\n- {跨产品引流/共享组件}\n```\n\n## 资源分配模型\n\n### 核心原则\nCEO一个人 + AI团队，资源 = **CEO时间 + Agent算力 + 注意力**。\n\n### 分配公式\n```\n产品资源占比 = (健康度得分 × 战略权重) / Σ(所有产品的 健康度×权重)\n```\n\n其中**战略权重**由CEO指定（1-3倍）：\n- 3× = CEO明确标注的战略重点\n- 2× = 有明确增长潜力\n- 1× = 维护型\n\n### 分配约束\n| 规则 | 约束 |\n|------|------|\n| 单产品上限 | ≤60%（避免all-in风险） |\n| 单产品下限 | ≥5%（维护最低限度） |\n| 新产品试验 | 独立预留10-15%（不从现有产品扣） |\n| 调整频率 | 月度review，CEO批准后生效 |\n\n### Agent工时分配（参考）\n| 角色 | 分配逻辑 |\n|------|---------|\n| fullstack-dev | 按产品资源占比分配coding时间 |\n| content-chief | 按产品资源占比分配内容产出 |\n| growth-lead | 聚焦当前AARRR最弱环节的产品 |\n| data-analyst | 各产品均匀覆盖（数据不能偏科） |\n| intel-analyst | 聚焦⭐明星和❓问号的竞品 |\n\n### 资源重分配触发条件\n| 信号 | 动作 |\n|------|------|\n| 产品健康度连续2月下降 | 减少投入，除非CEO坚持 |\n| 产品健康度突破4.0 | 考虑加大投入（明星候选） |\n| ❓问号产品3月未达标 | 降为🐕瘦狗，释放资源 |\n| CEO新增产品 | 从预留池分配，不挤占现有 |\n| 竞品重大威胁 | 临时增加相关产品投入（参谋长判断+CEO批准） |\n\n## 跨产品协同策略\n\n### 协同类型\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **流量协同** | A产品用户引导到B产品 | 影视站→工具站推荐 |\n| **技术协同** | 共享组件/服务 | 统一认证、共享UI库 |\n| **数据协同** | 共享用户画像/行为数据 | 用户兴趣标签跨产品复用 |\n| **品牌协同** | 统一品牌矩阵 | 主品牌→子品牌信任传递 |\n| **SEO协同** | 站群互链+内容交叉引用 | 但要注意搜索引擎惩罚风险 |\n\n### 协同评估\n每次发现协同机会时评估：\n- **收益**: 预期带来多少增量（流量/用户/收入）\n- **成本**: 需要多少开发/运营投入\n- **风险**: 是否有负面影响（如SEO惩罚、品牌稀释）\n- **依赖**: 需要几个Agent配合，协调成本高不高\n\n## 定价策略分析框架\n\n### 定价三角\n```\n        价值感知\n       /        \\\n      /          \\\n竞品价格 ——— 成本底线\n```\n\n### 定价方法选择\n| 阶段 | 方法 | 说明 |\n|------|------|------|\n| 冷启动 | 免费/Freemium | 先要用户，不要收入 |\n| 有留存 | 功能分级 | 基础免费 + 高级付费 |\n| 有品牌 | 价值定价 | 按用户获得的价值定价 |\n| 成熟期 | 竞品对标 | 参考竞品定价+差异化溢价 |\n\n### 价格测试框架\n- A/B测试不同价格点（需要足够流量）\n- 用户调研：\"你愿意为X付多少\"\n- 竞品定价对标分析（intel-analyst 提供数据）\n- Van Westendorp价格敏感度（如果能做调研的话）\n\nFile v2.0.4:references/agent-refs/content-chief/methodology.md\n\n# 内容方法论\n\n## 内容漏斗匹配（Content-Funnel Fit）\n\n每篇内容明确服务哪个用户旅程阶段：\n\n| 阶段 | 用户意图 | 内容类型 | 示例 |\n|------|---------|---------|------|\n| **TOFU** 认知 | \"有这类工具吗？\" | 科普、趋势、问题解答 | \"2026年影视数据库对比\" |\n| **MOFU** 考虑 | \"哪个适合我？\" | 对比评测、教程、案例 | \"Richlyva vs IMDb 角色图谱对比\" |\n| **BOFU** 决策 | \"我要用这个\" | 上手指南、API文档 | \"5分钟搭建影视宇宙\" |\n\n**比例约 5:3:2（TOFU:MOFU:BOFU）**\n\n## 内容日历与排期\n\n### 发布节奏（冷启动期）\n| 类型 | 频率 | 负责 | 说明 |\n|------|------|------|------|\n| 长文（>1500词） | 2篇/周 | 内容主编主导 | SEO/GEO核心资产 |\n| 短文/列表（500-1000词） | 2篇/周 | 可委托或模板化 | 长尾关键词覆盖 |\n| 社区帖（Reddit/HN/Quora） | 3条/周 | 配合growth-lead | 引流+品牌曝光 |\n| FAQ/更新日志 | 按需 | 配合product-lead | 站内SEO+用户信任 |\n\n### 选题流程（4步）\n1. **关键词池** → 从 SEO/GEO 关键词研究中选取本周目标词\n2. **意图匹配** → 判断搜索意图（信息型/导航型/交易型）→ 匹配漏斗阶段\n3. **竞品内容差距** → intel-analyst 竞品数据中找对手没覆盖的角度\n4. **排期** → 写入 `shared/knowledge/content-plan-{YYYY}-W{nn}.md`\n\n### 内容日历模板\n```markdown\n## 内容计划 {YYYY}-W{nn}\n| 日期 | 标题 | 漏斗 | 目标关键词 | 渠道 | 状态 |\n|------|------|------|-----------|------|------|\n| 周一 | ... | TOFU | ... | Blog | 草稿/发布/推迟 |\n```\n\n## SEO/GEO 关键词研究流程\n\n### 第1步：种子词收集\n- 产品核心功能词（character map, movie universe, film database）\n- 用户痛点词（从 data-analyst 反馈/搜索query）\n- 竞品品牌词+替代词（IMDb alternative, letterboxd vs）\n- AI搜索常见问法（best tool for..., how to find...）\n\n### 第2步：扩展与验证\n| 工具/方法 | 用途 | 成本 |\n|-----------|------|------|\n| Google Suggest / People Also Ask | 长尾扩展 | 免费 |\n| Bing Webmaster | 实际搜索数据 | 免费 |\n| 竞品页面H1/H2/Title扫描 | 发现覆盖差距 | 免费 |\n| Reddit/Quora热门问题 | 真实用户措辞 | 免费 |\n| Google Trends | 趋势验证 | 免费 |\n\n### 第3步：关键词评分\n| 因子 | 说明 | 权重 |\n|------|------|:----:|\n| 搜索量预估 | 高/中/低（无精确数据时用趋势判断） | 30% |\n| 竞争难度 | SERP前10是大站还是小站 | 25% |\n| 意图匹配 | 和产品功能的相关度 | 25% |\n| GEO潜力 | AI搜索是否会引用此类内容 | 20% |\n\n### 第4步：关键词档案\n```markdown\n## 关键词: {keyword}\n- 搜索量预估: 高/中/低\n- 竞争难度: 高/中/低\n- 意图: 信息型/导航型/交易型\n- 漏斗阶段: TOFU/MOFU/BOFU\n- SERP现状: {前3结果概述}\n- GEO机会: {AI搜索是否已有引用}\n- 优先级: P1/P2/P3\n- 计划内容: {标题草案}\n```\n\n## 内容效果评估框架\n\n### 发布后追踪（分阶段）\n| 时间点 | 检查项 | 合格线 |\n|--------|--------|--------|\n| 发布即时 | Google索引（site:命令） | 48h内收录 |\n| 7天 | Search Console曝光量 | >0（有曝光） |\n| 14天 | 排名位置 | 目标词进前50 |\n| 30天 | 点击量、平均排名 | 目标词进前20 |\n| 60天 | 自然流量占比 | 该页有稳定流量 |\n\n### 内容健康度分级\n| 等级 | 标准 | 行动 |\n|------|------|------|\n| 🟢 健康 | 排名稳定/上升，有流量 | 保持，考虑扩展内链 |\n| 🟡 待优化 | 排名11-30，有曝光无点击 | 优化标题/摘要/结构 |\n| 🔴 失败 | 60天无排名无流量 | 分析原因：选题错？内容薄？竞争强？ |\n| ⚫ 过时 | >6个月未更新且流量下降 | 更新或合并 |\n\n### 内容ROI评估\n```\n内容价值 = (自然流量 × 预估CPC) / 创作时间(小时)\n```\n无精确数据时用定性判断：这篇内容带来了注册/咨询/品牌曝光吗？\n\n## 内容复用策略（1→N）\n\n一篇核心长文可拆解为：\n\n| 衍生形式 | 渠道 | 改造量 |\n|---------|------|--------|\n| 精华摘要（300字） | Reddit/HN帖子 | 低 |\n| FAQ拆条 | Quora回答 | 低 |\n| 数据可视化 | Twitter/社媒图 | 中 |\n| 对比表格 | 独立落地页 | 中 |\n| 视频脚本大纲 | YouTube（未来） | 高 |\n| Newsletter片段 | 邮件（未来） | 低 |\n\n**规则：核心长文先发博客 → 等收录后 → 再拆分发各渠道（避免重复内容惩罚）**\n\n## 写作检查清单（发布前必检）\n- [ ] 标题含关键词且吸引点击\n- [ ] 开头30字回答搜索意图（AI搜索截取用）\n- [ ] H2/H3 含关键词变体\n- [ ] FAQ 区块 3-5 个（Schema+GEO友好）\n- [ ] 内链 ≥2 个\n- [ ] CTA 明确\n- [ ] 语言地道自然（面向全球英文用户）\n- [ ] 标注：目标关键词、漏斗阶段、目标AARRR环节\n- [ ] Meta description ≤155字符，含关键词+行动号召\n- [ ] 图片有alt text（含关键词变体）\n- [ ] URL slug 简短含关键词（英文短横线分隔）\n\nFile v2.0.4:references/agent-refs/data-analyst/methodology.md\n\n# 数据分析方法论\n\n## 反馈分类框架\n\n收到用户反馈后按以下维度分类，再写信给 product-lead：\n\n| 维度 | 分类 | 权重 |\n|------|------|:----:|\n| **类型** | Bug / 功能请求 / 体验抱怨 / 表扬 / 疑问 | — |\n| **频率** | 单次 / 偶发(2-5) / 高频(>5) | 高频×3 |\n| **情绪** | 愤怒 / 失望 / 中性 / 积极 | 愤怒×2 |\n| **用户价值** | 免费 / 付费 / KOL | 付费×2 |\n| **影响范围** | 个人 / 部分用户 / 全体 | 全体×3 |\n\n**写给 product-lead 必须包含：**\n1. 原始反馈（引用原文）\n2. 分类标签\n3. 同类反馈数量（\"过去7天X条类似\"）\n4. 你的分析（真需求还是个例？有 workaround？）\n\n**绝不直接写\"建议开发XX\"** — 你只提供数据+洞察，方案是 product-lead 的事。\n\n## 数据采集清单（按产品生命周期）\n\n### 冷启动期（上线前/刚上线，当前阶段）\n优先接入，成本从低到高：\n\n| 数据源 | 追踪指标 | 工具 | 成本 | 优先级 |\n|--------|---------|------|------|:------:|\n| 搜索引擎 | 收录数、排名、曝光、点击 | Google Search Console + Bing Webmaster | 免费 | P0 |\n| 网站流量 | UV/PV、来源、页面停留、跳出率 | Umami / Plausible (自部署) 或 GA4 | 免费 | P0 |\n| 用户行为 | 点击热图、滚动深度、关键路径 | Microsoft Clarity | 免费 | P1 |\n| 错误监控 | JS报错、API错误率、慢查询 | Sentry (免费额度) | 免费 | P1 |\n| 社媒提及 | 品牌词、竞品词、行业词 | 手动 + intel-analyst | 免费 | P2 |\n| 用户反馈 | 评分、评论、邮件 | 内置反馈widget / G2 | 免费 | P2 |\n\n### 增长期（有稳定流量后）\n| 数据源 | 追踪指标 | 工具 |\n|--------|---------|------|\n| 转化漏斗 | 注册转化率、激活率、付费转化 | 自建事件追踪 |\n| 留存 | D1/D7/D30 留存率 | 数据库查询 |\n| 收入 | ARPU、MRR、Churn | Stripe Dashboard |\n| A/B测试 | 变体胜率、置信度 | GrowthBook (免费) |\n\n### 核心指标仪表盘（上线后必建）\n```\n北极星指标: {待定义，如\"周活跃探索次数\"}\n├── 获客: 日新增UV, 搜索流量占比, 社媒引荐量\n├── 激活: 注册率, 首次搜索完成率, 首次收藏率\n├── 留存: D1/D7 回访率\n├── 收入: (暂缓)\n└── 推荐: 分享率, 外链引用数\n```\n\n## 报告模板\n\n### 周报模板\n```markdown\n# 数据周报 {YYYY}-W{nn}\n\n## 概览\n- 报告周期: {起止日期}\n- 数据源: {列出所有使用的数据源}\n- 数据完整性: ✅完整 / ⚠️部分缺失（注明）/ ❌无数据\n\n## 核心指标\n| 指标 | 本周 | 上周 | 环比 | 趋势 |\n|------|------|------|------|------|\n| UV | | | | ↑↓→ |\n| 注册数 | | | | |\n| 搜索次数 | | | | |\n\n## 流量来源分布\n| 来源 | 占比 | 变化 |\n|------|------|------|\n\n## 异常与预警\n- {异常描述 + 可能原因 + 建议关注}\n\n## 洞察（给 product-lead）\n1. {数据发现 → 含义 → 建议方向（不是方案）}\n\n## 下周关注\n- {需要重点追踪的指标或实验}\n```\n\n### 月报（在周报基础上追加）\n```markdown\n## 月度趋势\n- 本月 vs 上月 vs 上上月（3月对比）\n- 关键拐点和归因分析\n\n## 用户反馈月度汇总\n- 按分类框架统计Top 5主题\n- 情绪趋势变化\n\n## 实验结果汇总\n- 本月实验清单 + 结论（配合 growth-lead）\n```\n\n## 归因分析方法\n\n### 流量归因模型（简易版，适合冷启动）\n| 归因方式 | 适用场景 | 做法 |\n|---------|---------|------|\n| **UTM标签** | 所有主动推广链接 | 统一格式: `?utm_source=X&utm_medium=Y&utm_campaign=Z` |\n| **来源页面 Referrer** | 自然流量 | 分析 referrer 域名分布 |\n| **首次/末次触点** | 转化路径 | 记录用户首次来源 + 注册时来源 |\n| **搜索词** | SEO效果 | Search Console 查询报告 |\n\n### UTM命名规范\n```\nutm_source:   google / bing / reddit / hackernews / twitter / direct\nutm_medium:   organic / social / referral / email / community\nutm_campaign: {描述性名称，如 launch-week, reddit-ama, blog-character-maps}\n```\n\n### 内容→转化归因\n```\n内容贡献 = 该内容页作为入口的注册数 / 总注册数\n```\n定期输出\"内容贡献排行\"给 content-chief，指导内容策略。\n\n## 异常检测规则\n- 日活偏离7日均值 >20% → 预警\n- 关键转化率偏离 >15% → 预警\n- 新增差评 >3条/天 → 预警\n- 搜索排名核心词下降 >10位 → 预警（配合 Search Console）\n- 错误率 >1% → 预警（配合 Sentry）\n- 无数据时：标注\"无数据源，跳过\"，不编造\n\n## 数据呈现标准\n- 必须注明时间范围和数据源\n- 同比环比都给，趋势用 ↑↓→ 标注\n- 不编造数据，没有就说没有\n- 低置信度结论标注 `[低置信度]`\n- 图表优先于纯数字（能给表格就给表格）\n- 关键数字加粗，趋势附原因假设\n\nFile v2.0.4:references/agent-refs/fullstack-dev/coding-behavior-fallback.md\n\n# Coding Behavior (Fallback)\n\n> **This file is only needed when the coding-lead skill is NOT loaded.**\n> If coding-lead is loaded, it provides all these rules in more detail. Ignore this file.\n\n### Task Classification\n| Level | Criteria | Action |\n|-------|----------|--------|\n| Simple | Single file, <60 lines | Do directly with read/write/edit/exec |\n| Medium | 2-5 files, clear scope | Spawn Claude Code via ACP |\n| Complex | Architecture change, multi-module | Plan first, then spawn |\n\n### Context Injection\nBefore spawning Claude Code, gather and inject into prompt:\n1. **产品知识文件** (`shared/products/{product}/` — architecture.md, database.md, services.md 等相关文件)\n2. Project info (CLAUDE.md, README, package.json/composer.json)\n3. Coding standards (tech-standards.md)\n4. Past decisions (search memory for related work)\n5. Known pitfalls from memory\n\n### Prompt Structure for Claude Code\n```\n## Project\n- Path: [project dir]\n- Stack: [from docs]\n\n## Coding Standards\n[From tech-standards.md]\n\n## Historical Context\n[From memory]\n\n## Task\n[Description]\n\n## Acceptance Criteria\n- [ ] ...\n\nBefore finishing:\n1. Run linter if available, fix issues\n2. Run tests if available, ensure they pass\n3. Report results in final output\n\nWhen completely finished, run:\nopenclaw system event --text \"Done: [summary]\" --mode now\n```\n\n### Spawn Rules\n- Always set cwd to the project directory, NOT ~/.openclaw/ or workspace-team/\n- Independent tasks can run in parallel (2-3 sessions max)\n- Track via sessions_list\n- Never let coding agents modify files outside the project directory\n\n### QA Isolation (Critical)\n- QA tests must be spawned in a SEPARATE session from implementation\n- QA prompt gets requirements + interface definitions only, NOT implementation code\n\n### Review by Complexity\n- **Simple**: no review, works = done\n- **Medium**: quick check -- success + tests pass + no obvious errors\n- **Complex**: full checklist (logic, security, performance, style, tests)\n\n### Coding Roles (Complex Tasks Only)\n- **Architect**: system design, DB schema, API contracts\n- **Frontend**: UI components, state management\n- **Backend**: API endpoints, business logic\n- **Reviewer**: independent code review\n- **QA**: test writing, edge case analysis\n\nFlow: Research -> Plan -> Architect(spawn) -> Implement(spawn, can parallel) -> Review(spawn) -> Fix -> Record.\nSkip roles that don't apply. Simple/medium: no roles, single spawn.\n\n### Smart Retry (max 3)\n1. Analyze failure\n2. Rewrite prompt\n3. Retry improved\n4. Max 3 attempts → stop, report to chief-of-staff\n\n### Prompt Pattern Library\n- Record successful prompt structures in memory\n- Search memory for similar past tasks before spawning\n\n### Progress Updates\n- On start/completion/error: notify appropriately\n- Kill runaway sessions and report\n\n## Task Tracking\n\nTrack active coding tasks in `<project>/.openclaw/active-tasks.json`:\n- Register each spawned CC session with: id, task, branch, status, startedAt\n- Update on completion/failure\n- Check before spawning to avoid duplicates\n\n## Definition of Done\n\nMedium: CC success + lint pass + tests pass + no unrelated changes + logged in memory\nComplex: all of above + code review + QA tests + UI screenshots if applicable\n\n## UI Screenshot Rule\nIf a task changes visible UI, the completion report must describe visual changes.\n\nFile v2.0.4:references/agent-refs/fullstack-dev/deep-dive-protocol.md\n\n# Deep Dive Scan Protocol\r\n\r\n> Load on demand: only read when a scan request is received.\r\n\r\n## When to Scan\r\n- Product-lead sends a Deep Dive request via inbox\r\n- New product added to shared/products/_index.md\r\n- After major code changes (L4 incremental)\r\n- Periodic health checks (L3, requested by product-lead)\r\n\r\n## Execution Protocol\r\n\r\n1. **Read the request** from inbox: product name, code directory, scan level, focus areas\r\n2. **Enter the project directory** (read-only unless explicitly told to modify code)\r\n3. **Detect tech stack** automatically:\r\n   - `composer.json` → Laravel/PHP\r\n   - `package.json` → Node/React/Vue (check framework field, dependencies)\r\n   - `requirements.txt` / `pyproject.toml` → Python/Django/FastAPI\r\n   - `go.mod` → Go | `Cargo.toml` → Rust\r\n   - Multiple? Note it's a monorepo or full-stack project\r\n4. **Execute scan commands** per stack (see below)\r\n5. **Write knowledge files** to `shared/products/{product}/`\r\n6. **Log the scan** in `shared/products/{product}/changelog.md`\r\n7. **Notify product-lead** via inbox when done\r\n\r\n## Per-Stack Scan Strategies\r\n\r\n### Laravel/PHP\r\n- L0: tree, composer.json, .env.example\r\n- L1: migrations → database.md, route:list → routes.md + api.md, Models → models.md\r\n- L2: Services/Actions → services.md, Middleware/Policies → auth.md, Jobs/Listeners → jobs-events.md, Console/Kernel → scheduled tasks, config/services.php → integrations.md\r\n- L3: grep TODO/FIXME/HACK, test coverage, Dockerfile, CI, security scan\r\n\r\n### React/Vue/Frontend\r\n- L0: tree src/, package.json\r\n- L1: components → frontend.md, router config, state management stores\r\n- L2: API client layer → data-flow.md, form validation, auth/token handling, i18n\r\n\r\n### Python/Django/FastAPI\r\n- L1: models.py → database.md, urls.py → routes.md, serializers → api.md\r\n- L2: views.py → services.md, middleware/permissions → auth.md, celery → jobs-events.md\r\n\r\n### General (any stack)\r\n- git log --oneline -30, find *.md, grep TODO/FIXME/HACK, tree -L 3\r\n- .env.example → config-env.md, Dockerfile/CI → devops.md, test dirs → test-coverage.md\r\n\r\n## Content Quality Standards\r\n\r\nKnowledge files must capture:\r\n1. **Facts** — what exists (tables, routes, classes)\r\n2. **Relationships** — how things connect\r\n3. **Rationale** — why it's built this way\r\n4. **Implicit rules** — undocumented business logic buried in code\r\n5. **Gotchas** — surprises for a new developer\r\n6. **Cross-module coupling** — where changing A silently breaks B\r\n7. **Performance notes** — N+1 queries, missing indexes, bottlenecks\r\n\r\n## Output Format\r\n\r\n```\r\n# {Title} - {Product Name}\r\n> Auto-generated by the dual-dev Deep Dive flow.\r\n> Last scan: YYYY-MM-DD | Level: L{n} | Stack: {stack}\r\n> Code directory: {path}\r\n```\r\n\r\n## L4 Incremental Protocol\r\n\r\n1. `git log --oneline --since={last scan date}`\r\n2. `git diff --stat` → map changed files to knowledge files\r\n3. Read only changed files, update only affected knowledge files\r\n4. Log in changelog.md\r\n\r\n## Large Projects (>500 files)\r\nSpawn Claude Code for the scan: provide scan level, output directory, content standards. Read-only mode.\n\nFile v2.0.4:references/agent-refs/fullstack-dev/methodology.md\n\n# 工程方法论\n\n## 架构决策记录（ADR）\n\n每个重要技术决策必须留记录。\"重要\"的定义：影响 >1 个模块、不可轻易回退、或团队有分歧。\n\n### ADR 模板\n```markdown\n# ADR-{序号}: {决策标题}\n\n- 日期: {YYYY-MM-DD}\n- 状态: 提议 / 采纳 / 废弃 / 替代(被ADR-XX替代)\n- 决策者: product-lead + fullstack-dev\n\n## 背景\n{为什么需要做这个决策？当前的问题是什么？}\n\n## 决策\n{我们选择了什么方案？}\n\n## 备选方案\n| 方案 | 优点 | 缺点 | 排除原因 |\n|------|------|------|---------|\n| A | | | |\n| B(采纳) | | | — |\n| C | | | |\n\n## 影响\n- 影响模块: {列出}\n- 迁移成本: 高/中/低/无\n- 风险: {可能的问题}\n- 技术债: {引入了什么债/减少了什么债}\n\n## 后续\n- [ ] {需要做的事}\n```\n\n### ADR 存放位置\n`shared/products/{产品}/adrs/` 目录，或写在 architecture.md 的 ## 决策记录 章节。\n\n### 触发 ADR 的场景\n- 新增/替换框架或主要依赖\n- 数据库 schema 重大变更\n- API 破坏性变更\n- 新增微服务/模块拆分\n- 认证/授权方案变更\n- 部署架构变更\n\n## 代码 Review 清单\n\n### 通用检查（所有PR）\n- [ ] **功能正确**: 满足验收标准，边界情况处理\n- [ ] **无回退**: 不破坏已有功能（跑现有测试）\n- [ ] **命名清晰**: 变量/函数/类名可读、一致\n- [ ] **无硬编码**: 配置项走 env/config，不写死\n- [ ] **错误处理**: try-catch 合理，用户友好错误提示\n- [ ] **日志合理**: 关键操作有日志，无敏感信息泄露\n\n### 安全检查\n- [ ] **输入验证**: 所有用户输入已校验/转义\n- [ ] **SQL注入**: 使用参数化查询/ORM，无拼接SQL\n- [ ] **XSS**: 输出已转义，dangerouslySetInnerHTML 有理由\n- [ ] **CSRF**: 表单有token\n- [ ] **认证/授权**: 路由权限检查，无越权访问\n- [ ] **敏感数据**: 密码哈希、token不明文、日志无敏感信息\n- [ ] **依赖安全**: 无已知漏洞依赖（npm audit / composer audit）\n\n### 性能检查\n- [ ] **N+1查询**: 列表页是否 eager load\n- [ ] **索引**: 新增查询条件是否有对应索引\n- [ ] **大数据**: 分页是否正确，无全表扫描\n- [ ] **缓存**: 高频读取是否命中缓存\n- [ ] **前端**: 组件是否合理 memo/懒加载，bundle size\n\n### 可维护性检查\n- [ ] **单一职责**: 函数/组件不超过 100 行（超了要拆）\n- [ ] **重复代码**: DRY，有无可抽取的公共逻辑\n- [ ] **测试**: 新功能有测试，改动不破坏已有测试\n- [ ] **文档**: 复杂逻辑有注释，API 变更更新文档\n- [ ] **兼容性**: 数据库 migration 可回滚\n\n### Review 严格度（按任务复杂度）\n| 复杂度 | Review 方式 |\n|--------|------------|\n| Simple | 自检通过即可（功能ok + lint pass + tests pass） |\n| Medium | 过通用 + 安全清单 |\n| Complex | 全部清单 + 架构影响评估 + 单独QA |\n\n## 技术债评估框架\n\n### 技术债分类\n| 类型 | 描述 | \n\nArchive v2.0.2: 18 files, 55189 bytes\n\nFiles: README.md (4997b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (14240b), references/soul-templates.md (19800b), scripts/deploy.js (21405b), SKILL.md (24336b), team-builder.json (394b), _meta.json (131b)\n\nArchive v2.0.1: 18 files, 55442 bytes\n\nFiles: README.md (5535b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (14240b), references/soul-templates.md (19800b), scripts/deploy.js (21405b), SKILL.md (24336b), team-builder.json (394b), _meta.json (131b)\n\nArchive v2.0.0: 18 files, 54384 bytes\n\nFiles: README.md (4992b), references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/coding-behavior-fallback.md (1234b), references/shared-templates.md (13729b), references/soul-templates.md (19702b), scripts/deploy.js (19707b), SKILL.md (24282b), team-builder.json (394b), _meta.json (131b)\n\nArchive v1.0.32: 16 files, 49720 bytes\n\nFiles: references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3170b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/shared-templates.md (13550b), references/soul-templates.md (19482b), scripts/deploy.js (15914b), SKILL.md (24138b), team-builder.json (394b), _meta.json (132b)\n\nArchive v1.0.31: 15 files, 48301 bytes\n\nFiles: references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3096b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/shared-templates.md (13091b), references/soul-templates.md (18359b), scripts/deploy.js (15456b), SKILL.md (22851b), _meta.json (132b)\n\nArchive v1.0.30: 15 files, 48072 bytes\n\nFiles: references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3096b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/shared-templates.md (13091b), references/soul-templates.md (18359b), scripts/deploy.js (15141b), SKILL.md (22677b), _meta.json (132b)\n\nArchive v1.0.24: 15 files, 54819 bytes\n\nFiles: references/agent-refs/chief-of-staff/dashboard-template.md (759b), references/agent-refs/chief-of-staff/strategy-methodology.md (4616b), references/agent-refs/content-chief/methodology.md (5071b), references/agent-refs/data-analyst/methodology.md (4895b), references/agent-refs/fullstack-dev/coding-behavior-fallback.md (3343b), references/agent-refs/fullstack-dev/deep-dive-protocol.md (3096b), references/agent-refs/fullstack-dev/methodology.md (5355b), references/agent-refs/growth-lead/methodology.md (1669b), references/agent-refs/intel-analyst/methodology.md (1564b), references/agent-refs/product-lead/methodology.md (2062b), references/shared-templates.md (12892b), references/soul-templates.md (18443b), scripts/deploy.js (35086b), SKILL.md (19812b), _meta.json (132b)","readmeExcerpt":"Skill: Team Builder Owner: beyound87 Summary: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Tags: latest:3.0.0 Version history: v3.0.0 | 2026-04-02T15:29:56.429Z | user 不使用acp session模式来模拟claude持久化 v2.0.6 | 2026-03-30T00:50:15.771Z | user 不使用acp session模式来模拟claude持久化 v2.0.4 | 2026-03-29T18:59:53.219Z | user 不使用a","codeSnippets":[],"executableExamples":[{"language":"bash","snippet":"node -v        # 需要 Node.js 16+\nopenclaw status  # 确认 OpenClaw 已运行"},{"language":"bash","snippet":"node ~/.openclaw/skills/team-builder/scripts/deploy.js"},{"language":"bash","snippet":"node <工作区路径>/apply-config.js"},{"language":"powershell","snippet":"powershell <工作区路径>/create-crons.ps1"},{"language":"bash","snippet":"bash <工作区路径>/create-crons.sh"},{"language":"bash","snippet":"openclaw gateway restart"}],"parameters":null,"dependencies":[],"permissions":[],"extractedFiles":[{"path":"SKILL.md","content":"---\r\nname: team-builder\r\ndescription: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。\r\n---\r\n\r\n# Team Builder\r\n\r\n一条命令部署完整多 Agent 团队骨架，包含角色、信箱、看板、产品知识目录、cron 巡检、双开发轨模板。\r\n\r\n> 详细安装步骤和首次配置引导见 `README.md`（中文完整版）。\r\n\r\n## 前置条件\r\n\r\n- OpenClaw 已安装并运行\r\n- Node.js 16+\r\n- 已配置至少一个模型 provider\r\n\r\n> **此技能会修改系统配置**：`apply-config.js` 会写入 `openclaw.json`，`create-crons.*` 会创建 cron 任务。均需手动执行，不自动运行。\r\n\r\n### Internal Dispatch Protocol (MANDATORY)\r\n- **Standard agents** (product-lead, growth-lead, intel-analyst, devops, etc.): `sessions_spawn(runtime=\"subagent\", mode=\"run\")` — **禁止带 `streamTo`**\r\n- **ACP agents** (fullstack-dev): `sessions_spawn(runtime=\"acp\")` — 可带 `streamTo=\"parent\"`\r\n- 结果通过 auto-announce 推送；不要轮询 `sessions_list` 或 `subagents list`\r\n- chief-of-staff 是群内唯一入口；其他 agent 内部 spawn，不直接监听群聊\r\n- 详见 `shared/knowledge/team-workflow.md` 零章\r\n\r\n### 参谋长执行边界（MANDATORY）\r\n- **参谋长不下地干活**：任何多步骤任务（编码、调研、分析、内容、部署）→ 全部派给对应 agent\r\n- **参谋长不亲自开子代理干活**：spawn 子代理执行具体业务 = 等价于自己干活\r\n- **参谋长做且只做**：任务拆解（L1-L4 复杂度判断）、编排、派活、监控、结果汇总\r\n- 先执行后汇报；默认输出只保留：已做什么 / 拿到什么结果 / 卡在哪里\r\n\r\n### 子代理使用规则（MANDATORY，全团队适用）\r\n- **优先主 agent 自己干**：干得过来时不开子代理\r\n- **分析透再派**：开子代理前必须先分析清楚边界/依赖/输入输出\r\n- **子代理只做原子任务**：一句话说清、执行完就结束，不做判断或策略决策\r\n- **主 agent 全权负责**：判断、策略、经验积累——子代理不做这三件\r\n- **子代理结果由主 agent 汇总**后再上报参谋长\r\n\r\n### Credentials involved\r\n- **Telegram bot tokens** (optional) -- stored in openclaw.json, used for agent-to-Telegram binding\r\n- **Model API keys** -- must already be configured in your OpenClaw model providers (not handled by this skill)\r\n\r\n### Recommended\r\n- Review generated `apply-config.js` before running\r\n- Check the backup of openclaw.json after running\r\n- Test with 2-3 agents before enabling all cron jobs\r\n\r\n## Team Architecture\r\n\r\nDefault reference architecture for a SaaS/growth multi-agent team (customizable to 2-10 agents):\r\n\r\n```\r\nCEO\r\n |-- Chief of Staff (dispatch + strategy + efficiency)\r\n |-- Data Analyst (data + user research)\r\n |-- Growth Lead (GEO + SEO + community + social media)\r\n |-- Content Chief (strategy + writing + copywriting + i18n)\r\n |-- Intel Analyst (competitor monitoring + market trends)\r\n |-- Product Lead (product management + tech architecture)\r\n |-- DevOps (delivery / deploy / environment / acceptance)\r\n |-- Fullstack Dev (implementation / module deep dive / ACP coding session)\r\n```\r\n\r\n### Multi-Team Support\r\n\r\nOne OpenClaw instance can run multiple teams:\r\n\r\n```bash\r\nnode <skill-dir>/scripts/deploy.js                  # default team\r\nnode <skill-dir>/scripts/deploy.js --team alpha      # named team \"alpha\"\r\nnode <skill-dir>/scripts/deploy.js --team beta       # named team \"beta\"\r\n```\r\n\r\nNamed teams use prefixed agent IDs (`alpha-chief-of-staff`, `beta-growth-lead`) to avoid conflicts. Each team gets its own workspace subdirectory.\r\n\r\n### Flexible Team Size\r\n\r\nThe wizard lets you select 2-10 agents from the available roles. Skip roles you don't need. The 8-agent default covers mos"},{"path":"README.md","content":"# team-builder\n\n多 Agent 团队工作区生成工具。一条命令搭起完整的团队骨架，包含角色、信箱、看板、产品知识目录、cron 巡检、双开发轨模板。\n\n---\n\n## 🚀 首次使用：完整安装流程\n\n### 第一步：确认前置条件\n\n```bash\nnode -v        # 需要 Node.js 16+\nopenclaw status  # 确认 OpenClaw 已运行\n```\n\n### 第二步：运行生成脚本（交互式）\n\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js\n```\n\n脚本会依次询问：\n- 团队名称（如 `Alpha Team`）\n- 工作区路径（默认 `~/.openclaw/workspace-team`）\n- 时区（默认 `Asia/Shanghai`）\n- 晨报/晚报时间（默认 8 / 18）\n- 思考型模型（用于 chief-of-staff / product-lead / growth-lead）\n- 执行型模型（用于 devops / fullstack-dev 等）\n- CEO 称呼（如 `老板` / `Boss`）\n- 各角色名称（可直接回车用默认值）\n\n> 💡 不想交互？用 `--config`，见下方「非交互部署」。\n\n### 第三步：应用配置（写入 openclaw.json）\n\n```bash\nnode <工作区路径>/apply-config.js\n```\n\n这一步把生成的 Agent 写入 OpenClaw 配置，**必须执行**。\n\n### 第四步：创建定时任务\n\n**Windows（PowerShell）：**\n```powershell\npowershell <工作区路径>/create-crons.ps1\n```\n\n**macOS / Linux（Bash）：**\n```bash\nbash <工作区路径>/create-crons.sh\n```\n\n### 第五步：重启 Gateway\n\n```bash\nopenclaw gateway restart\n```\n\n### 第六步：CEO 填写项目上下文 ⚠️ 重要\n\n打开工作区里的 `shared/onboarding.md`，**填写以下内容**：\n\n```markdown\n## 当前项目/产品\n（是什么产品，目标用户，核心功能一句话）\n\n## 现在进展到哪步\n（已完成什么，当前在做什么）\n\n## 最近完成的里程碑\n（近期上线/交付了什么，时间）\n\n## 当前最大阻塞\n（卡在哪里，需要什么才能继续）\n\n## 技术栈/代码目录\n（主要技术栈，代码在哪个路径）\n```\n\n> ⚠️ **这一步非常重要**：所有 Agent 启动时会先读这个文件。不填的话 Agent 不知道你在做什么项目，会从零开始问你。\n> 填完后告知 chief-of-staff（直接在群里说一声），它会开始正式工作。\n\n### 第七步：验证（可选但推荐）\n\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --verify --config <配置文件路径>\n```\n\n输出结构化 JSON 检查报告，覆盖 `core / roles / fallback / workflow / hygiene / cron`。\n\n---\n\n## ⚙️ 非交互部署（推荐用于重复/CI 场景）\n\n准备配置文件 `team-builder.json`：\n\n```json\n{\n  \"teamName\": \"Alpha Team\",\n  \"workspaceDir\": \"~/.openclaw/workspace-team\",\n  \"timezone\": \"Asia/Shanghai\",\n  \"morningHour\": 8,\n  \"eveningHour\": 18,\n  \"thinkingModel\": \"claude-sonnet-4-5\",\n  \"executionModel\": \"claude-sonnet-4-5\",\n  \"ceoTitle\": \"老板\",\n  \"roles\": [\"chief-of-staff\", \"product-lead\", \"devops\", \"fullstack-dev\", \"growth-lead\", \"content-chief\", \"intel-analyst\", \"data-analyst\"],\n  \"roleNames\": {\n    \"chief-of-staff\": \"参谋长\",\n    \"product-lead\": \"产品总监\",\n    \"devops\": \"交付总监\",\n    \"fullstack-dev\": \"全栈工程师\",\n    \"growth-lead\": \"增长总监\",\n    \"content-chief\": \"内容主编\",\n    \"intel-analyst\": \"情报分析师\",\n    \"data-analyst\": \"数据分析师\"\n  }\n}\n```\n\n运行：\n```bash\nnode ~/.openclaw/skills/team-builder/scripts/deploy.js --config team-builder.json\n```\n\n---\n\n## 📁 生成物说明\n\n| 文件/目录 | 说明 |\n|-----------|------|\n| `AGENTS.md` | 启动引导 + 角色对照表 |\n| `SOUL.md` | 团队价值观（可自定义） |\n| `USER.md` | CEO 信息 |\n| `shared/onboarding.md` | **CEO 填写项目背景**（新增，首次必填） |\n| `shared/status/team-dashboard.md` | 团队状态面板（chief-of-staff 维护） |\n| `shared/decisions/active.md` | CEO 决策与指令 |\n| `shared/products/_index.md` | 产品矩阵目录 |\n| `shared/inbox/to-[role].md` | 各角色信箱 |\n| `shared/knowledge/team-workflow.md` | 团队工作流规范 |\n| `agents/[role]/SOUL.md` | 各角色身份与职责 |\n| `agents/[role]/MEMORY.md` | 各角色长期记忆 |\n| `apply-config.js` | 写入 openclaw.json（手动执行） |\n| `create-crons.ps1 / .sh` | 创建 cron 定时任务（手动执行） |\n| `references/coding-behavior-fallback.md` | fullstack-dev 编码行为兜底 |\n\n---\n\n## 🧠 Agent 启动顺序（已内置于 AGENTS.md）\n\n每个 Agent 启动时自动按此顺序读文件："},{"path":"_meta.json","content":"{\n  \"ownerId\": \"kn761rne5jajzfg0d8h6nge4hh8220cq\",\n  \"slug\": \"team-builder\",\n  \"version\": \"3.0.0\",\n  \"publishedAt\": 1775143796429\n}"},{"path":"references/agent-refs/chief-of-staff/dashboard-template.md","content":"# 团队面板模板\n\n```markdown\n# 团队状态面板\n> 参谋长维护 | 最后更新: YYYY-MM-DD HH:MM\n\n## 🔴 紧急/阻塞\n- [描述] — 涉及: [agent] — 原因: [原因] — 已持续: [时长]\n\n## 📊 各agent状态\n| Agent | 上次活跃 | 当前任务 | 状态 |\n|-------|---------|---------|------|\n| 参谋长 | 刚才 | 面板+简报 | ✅ |\n| 产品总监 | | | |\n| 增长总监 | | | |\n| 内容主编 | | | |\n| 数据分析 | | | |\n| 情报分析 | | | |\n| 全栈工程 | | | |\n\n状态: ✅正常 🔄进行中 🔴阻塞 ⏳等待 💤失联(>48h)\n\n## 📬 未处理信箱\n| 消息 | 发→收 | 优先级 | 已等待 | 状态 |\n\n## 🔗 跨agent任务链\n- [任务]: A(✅) → B(🔄) → C(⏳)\n\n## 📅 今日/明日重点\n1. ...\n```"},{"path":"references/agent-refs/chief-of-staff/strategy-methodology.md","content":"# 矩阵策略方法论\n\n## 产品组合评估框架\n\n### 产品健康度评分卡（每产品，月度更新）\n\n| 维度 | 指标 | 权重 | 评分(1-5) | 说明 |\n|------|------|:----:|:---------:|------|\n| **流量** | 月UV增长率 | 20% | | >20%=5, 10-20%=4, 0-10%=3, 持平=2, 下降=1 |\n| **留存** | D7回访率 | 20% | | >40%=5, 30-40%=4, 20-30%=3, 10-20%=2, <10%=1 |\n| **变现** | 收入/利润趋势 | 15% | | 盈利=5, 接近盈亏=4, 有收入=3, 无收入=2, 亏损=1 |\n| **竞争** | 差异化强度 | 15% | | 独占优势=5, 明显差异=4, 有差异=3, 同质=2, 落后=1 |\n| **效率** | 投入产出比 | 15% | | 高ROI=5, 正向=4, 中等=3, 低=2, 负=1 |\n| **战略** | 对矩阵协同价值 | 15% | | 核心枢纽=5, 高协同=4, 有协同=3, 独立=2, 拖累=1 |\n\n**总分 = Σ(评分 × 权重)**\n\n### 产品分类（简化BCG矩阵）\n\n| 类型 | 条件 | 策略 |\n|------|------|------|\n| ⭐ **明星** | 高增长 + 高份额（或潜力） | 全力投入，抢市场 |\n| 🐄 **现金牛** | 低增长 + 高份额 | 维护为主，收割利润，少投入 |\n| ❓ **问号** | 高增长 + 低份额 | 限期验证：3个月达标→升星，否则降级 |\n| 🐕 **瘦狗** | 低增长 + 低份额 | 最低维护或考虑下线，资源释放 |\n\n### 产品组合仪表盘模板\n```markdown\n# 产品矩阵健康度 — {YYYY}-{MM}\n\n| 产品 | 类型 | 健康度 | 本月重点 | 资源分配 |\n|------|------|:------:|---------|:--------:|\n| Richlyva | ❓问号 | 3.2/5 | 冷启动验证 | 40% |\n| ... | | | | |\n\n## 矩阵级风险\n- {跨产品风险}\n\n## 协同机会\n- {跨产品引流/共享组件}\n```\n\n## 资源分配模型\n\n### 核心原则\nCEO一个人 + AI团队，资源 = **CEO时间 + Agent算力 + 注意力**。\n\n### 分配公式\n```\n产品资源占比 = (健康度得分 × 战略权重) / Σ(所有产品的 健康度×权重)\n```\n\n其中**战略权重**由CEO指定（1-3倍）：\n- 3× = CEO明确标注的战略重点\n- 2× = 有明确增长潜力\n- 1× = 维护型\n\n### 分配约束\n| 规则 | 约束 |\n|------|------|\n| 单产品上限 | ≤60%（避免all-in风险） |\n| 单产品下限 | ≥5%（维护最低限度） |\n| 新产品试验 | 独立预留10-15%（不从现有产品扣） |\n| 调整频率 | 月度review，CEO批准后生效 |\n\n### Agent工时分配（参考）\n| 角色 | 分配逻辑 |\n|------|---------|\n| fullstack-dev | 按产品资源占比分配coding时间 |\n| content-chief | 按产品资源占比分配内容产出 |\n| growth-lead | 聚焦当前AARRR最弱环节的产品 |\n| data-analyst | 各产品均匀覆盖（数据不能偏科） |\n| intel-analyst | 聚焦⭐明星和❓问号的竞品 |\n\n### 资源重分配触发条件\n| 信号 | 动作 |\n|------|------|\n| 产品健康度连续2月下降 | 减少投入，除非CEO坚持 |\n| 产品健康度突破4.0 | 考虑加大投入（明星候选） |\n| ❓问号产品3月未达标 | 降为🐕瘦狗，释放资源 |\n| CEO新增产品 | 从预留池分配，不挤占现有 |\n| 竞品重大威胁 | 临时增加相关产品投入（参谋长判断+CEO批准） |\n\n## 跨产品协同策略\n\n### 协同类型\n| 类型 | 描述 | 示例 |\n|------|------|------|\n| **流量协同** | A产品用户引导到B产品 | 影视站→工具站推荐 |\n| **技术协同** | 共享组件/服务 | 统一认证、共享UI库 |\n| **数据协同** | 共享用户画像/行为数据 | 用户兴趣标签跨产品复用 |\n| **品牌协同** | 统一品牌矩阵 | 主品牌→子品牌信任传递 |\n| **SEO协同** | 站群互链+内容交叉引用 | 但要注意搜索引擎惩罚风险 |\n\n### 协同评估\n每次发现协同机会时评估：\n- **收益**: 预期带来多少增量（流量/用户/收入）\n- **成本**: 需要多少开发/运营投入\n- **风险**: 是否有负面影响（如SEO惩罚、品牌稀释）\n- **依赖**: 需要几个Agent配合，协调成本高不高\n\n## 定价策略分析框架\n\n### 定价三角\n```\n        价值感知\n       /        \\\n      /          \\\n竞品价格 ——— 成本底线\n```\n\n### 定价方法选择\n| 阶段 | 方法 | 说明 |\n|------|------|------|\n| 冷启动 | 免费/Freemium | 先要用户，不要收入 |\n| 有留存 | 功能分级 | 基础免费 + 高级付费 |\n| 有品牌 | 价值定价 | 按用户获得的价值定价 |\n| 成熟期 | 竞品对标 | 参考竞品定价+差异化溢价 |\n\n### 价格测试框架\n- A/B测试不同价格点（需要足够流量）\n- 用户调研：\"你愿意为X付多少\"\n- 竞品定价对标分析（intel-analyst 提供数据）\n- Van Westendorp价格敏感度（如果能做调研的话）"}],"languages":[],"docsSourceLabel":"CLAWHUB","editorialOverview":"在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Skill: Team Builder Owner: beyound87 Summary: 在 OpenClaw 上一键部署多 Agent SaaS 团队工作区。内置双开发轨（devops 交付 + fullstack-dev 实现）、实时 spawn 调度、cron 巡检、Deep Dive 产品知识目录、onboarding 引导。支持自定义角色、模型、时区，可选 Telegram 接入。 Tags: latest:3.0.0 Version history: v3.0.0 | 2026-04-02T15:29:56.429Z | user 不使用acp session模式来模拟claude持久化 v2.0.6 | 2026-03-30T00:50:15.771Z | user 不使用acp session模式来模拟claude持久化 v2.0.4 | 2026-03-29T18:59:53.219Z | user 不使用a","editorialQuality":{"score":100,"threshold":65,"status":"ready","wordCount":820,"uniquenessScore":54,"reasons":[]}},"media":{"evidence":{"source":"no-media","verified":false,"confidence":"low","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":"No screenshots, media assets, or demo links are available."},"primaryImageUrl":null,"mediaAssetCount":0,"assets":[],"demoUrl":null},"ownerResources":{"evidence":{"source":"unclaimed","verified":false,"confidence":"low","updatedAt":"2026-10-09T23:45:18.082Z","emptyReason":"This page has not been claimed by the agent owner."},"hasCustomPage":false,"customPageUpdatedAt":null,"customLinks":[],"structuredLinks":{"docsUrl":null,"demoUrl":null,"supportUrl":null,"pricingUrl":null,"statusUrl":null},"customPage":null},"relatedAgents":{"evidence":{"source":"protocol-neighbors","verified":false,"confidence":"medium","updatedAt":"2026-10-10T05:17:51.521Z","emptyReason":null},"items":[{"id":"8ebccd8e-3863-4187-8355-c3f14e1f9edf","entityType":"agent","canonicalPath":"/agent/iofficeai-aionui","slug":"iofficeai-aionui","name":"AionUi","description":"Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!","url":"https://github.com/iOfficeAI/AionUi","homepage":"https://www.aionui.com","source":"GITHUB_REPOS","protocols":["MCP","OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-10-09T19:11:12.944Z","createdAt":"2026-02-25T03:38:16.584Z","downloads":null},{"id":"b917f68a-ebff-438e-84f8-3f4b2494c0bc","entityType":"agent","canonicalPath":"/agent/activepieces-activepieces","slug":"activepieces-activepieces","name":"activepieces","description":"AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents","url":"https://github.com/activepieces/activepieces","homepage":"https://www.activepieces.com","source":"GITHUB_REPOS","protocols":["OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-04-15T02:22:12.426Z","createdAt":"2026-02-25T03:38:12.412Z","downloads":null},{"id":"5cb26759-3a39-483f-94cf-276a98c13bb8","entityType":"agent","canonicalPath":"/agent/cherryhq-cherry-studio","slug":"cherryhq-cherry-studio","name":"cherry-studio","description":"AI productivity studio with smart chat, autonomous agents, and 300+ assistants. Unified access to frontier LLMs","url":"https://github.com/CherryHQ/cherry-studio","homepage":"https://cherry-ai.com","source":"GITHUB_REPOS","protocols":["MCP","OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-04-11T14:38:40.986Z","createdAt":"2026-02-25T03:38:19.379Z","downloads":null},{"id":"6f6582d0-5d76-4f0f-b81d-86520247950b","entityType":"agent","canonicalPath":"/agent/copilotkit-copilotkit","slug":"copilotkit-copilotkit","name":"CopilotKit","description":"The Frontend for Agents & Generative UI. React + Angular","url":"https://github.com/CopilotKit/CopilotKit","homepage":"https://docs.copilotkit.ai","source":"GITHUB_REPOS","protocols":["OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-03-25T09:50:57.846Z","createdAt":"2026-02-25T03:39:14.617Z","downloads":null}],"links":{"hub":"/agent","source":"/agent/source/clawhub","protocols":[{"label":"OpenClaw","href":"/agent/protocol/openclew"}]}}}