90 天內建立 AI agent 團隊，系統性提升品牌市佔率
github.com/chihao919/oteams (private)blog/| Team | 輸入 | 輸入格式 | 產出 | 產出格式 |
|---|---|---|---|---|
| Website Analyzer | 網站 URL | https://example.com |
審計報告 + 缺口清單 + 團隊規格 | JSON + Markdown |
| Blog Writer Fleet | 文章題目 | slug + topic keywords |
EN + ZH + AEO 文章 | HTML (content-plan staging) |
| Website Mirror | 來源網站 URL | https://example.com |
鏡像文章 + 對應表 | HTML + source-map.json |
| Peak Experience | HubSpot + Gmail | CRM deals + email threads |
旅程地圖 + email 模板 + 話術 | Markdown + HubSpot templates |
| Bizbase Upgrade | 客戶來信 | Gmail thread content |
智能回信草稿 | Gmail draft (pre-filled) |
# 分析一個網站 /website-analyzer --url "https://watersonusa.com" # 寫一篇新文章 /blog-writer-fleet --topic "fire door spring hinge" --slug "fire-door-spring-hinge-guide" # 鏡像一個網站 /website-mirror --source "https://watersonusa.com" --target "/mirror/"
| Agent | Advances Leader O? | Unique Value? |
|---|---|---|
| 0. Leader | — (IS the O) | Guardian — accepts/rejects deliverables based on whether they create a genuine "different" feeling |
| 1. Journey Mapper | YES — without knowing where Peaks and Valleys currently are, no agent can design a "different" experience; this agent provides the coordinate system for O | YES — only agent that synthesizes HubSpot CRM + Gmail thread data into a unified journey map |
| 2. Win/Loss Analyst | YES — identifying which specific Peak moments drove wins directly informs what "different" means to each persona, making O actionable | YES — only agent applying MEDDICC framework to deal-level evidence; Journey Mapper sees the journey, Win/Loss sees the causation |
| 3. Peak Designer | YES — Peak Designer is the creative core of O: it translates data into the specific emotional design that makes Waterson "different" vs. product-only competitors | YES — only agent doing experience design (creative, not analytical); creates the HubSpot workflow implementation path per touchpoint |
| 4. Revival Strategist | YES — dormant clients who once felt Waterson was "different" can be reactivated by reminding them of that difference; revival directly re-extends O's reach | YES — REACTIVE scope only (prior interaction history); never overlaps with Agent 5's proactive work |
| 5. Cold Outreach Architect | YES — first impression IS the peak for cold prospects; this agent creates the "aha moment" the O describes as the first feeling of "different" | YES — PROACTIVE scope only (zero prior history); creates first-touch experiences for personas who have never interacted with Waterson |
| 6. Script Writer | YES — strategy without execution materials cannot create O's promised feeling; Script Writer is the final mile that makes the "different" feeling tangible | YES — only agent producing bilingual (EN+ZH) deployable materials; all other agents produce strategies, this agent produces deployable artifacts |
| 7. Performance Monitor | YES — O is not a one-time event; Performance Monitor ensures "different" keeps working over time and improves when metrics drop | YES — only agent measuring post-deployment reality vs. O promises; closes the feedback loop that prevents O from drifting |
| MOT 階段 | Elevation（提升） | Insight（洞察） | Pride（驕傲） | Connection（連結） |
|---|---|---|---|---|
| 進店 Awareness | Sample kit unboxing | AIA CEU aha moment | — | — |
| 轉化 Conversion | Ultra-fast quote (<2hr) | Fire code impact education | — | Reference call |
| 複購 Repurchase | VIP preview | New product briefing | Project completion celebration | Quarterly check-in |
| 推薦 Referral | — | — | Case study co-publish | CEO dinner |
| Persona | Primary Job | Pain When Hiring Wrong | Waterson's Unique Fit |
|---|---|---|---|
| Architect | Specify hardware that passes inspection | 60% fire doors fail → E&O liability | Compliance pre-check + AIA CEU |
| Contractor | Install fast, correctly, no callbacks | Spring hinge re-adjustment 6-18 mo | Hydraulic K51M = install-and-forget |
| Distributor | Reliable products, strong margin, low returns | Customer complaints about spring fatigue | K51M lifetime reliability + support |
| Agent | Advances Leader O? | Unique Value? |
|---|---|---|
| 0. Leader | — (IS the O) | Guardian — accepts/rejects capability upgrades based on whether they make replies feel expert and personalized |
| 1. Wiki Knowledge Indexer | YES — without relevant knowledge retrieval, the reply cannot demonstrate "understands my problem"; the index is the knowledge foundation that makes expertise possible | YES — only agent that processes all 67+ articles + product facts into a queryable index; no other agent does knowledge organization |
| 2. Shipping Calculator | YES — shipping quote in the reply removes friction and signals "we've already thought through your logistics," reinforcing the "understands my problem" feeling | YES — only agent handling pricing/shipping logic; pure computation with no overlap with knowledge or writing agents |
| 3. Reply Rewriter | YES — this agent IS the direct mechanism of O: it transforms generic quotes into expert replies with Challenger insight openings, directly creating the "understands my problem" reaction | YES — only agent that synthesizes all inputs (knowledge + shipping + marketing target + Peak-End) into the final customer-facing reply |
| 4. Extension Integrator | YES — O cannot be achieved if sales reps don't use the system; zero-friction UI integration ensures every reply benefits from the upgrade, making O scalable | YES — only agent handling Chrome Extension code and UI/UX; makes the O accessible in the tool sales reps already use every day |
| 客戶信件中的信號 | 行銷對象 | 回信語氣範例 |
|---|---|---|
| 提到 welded/barrel hinge, no mortise | Door Manufacturer (OEM) | "Your door manufacturer should specify mortise hinges..." |
| 提到 specific project, architect involved | Architect/Specifier | "For your spec, we recommend..." |
| 提到 PO number, account, reorder | Distributor | "Your distributor pricing for..." |
| 提到 DIY, home, personal project | End User | "For your door, we recommend..." |
| 提到 gate, pool, fence | Gate/Fence End User | "For your gate application..." |
| 無明確信號 | End User (default) | "For your application, we recommend..." |
| Metric | Target | Source |
|---|---|---|
| AI 爬蟲訪問量 | 週增 20% | Vercel Analytics bot filter |
| SEO 自然流量 | 月增 15% | Google Search Console |
| AI Citation | 月 ≥5 | scripts/check-citations.py |
| Article QG | ≥ 8.0 | 3-Model blind eval |
| P0 citation | ≥1/article | Claim Bank backcheck |
| # | NOT | INSTEAD |
|---|---|---|
| 1 | 三個策略群組各做各的 | 都服務同一個 O/G |
| 2 | 只寫文章不做缺口分析 | S2 分析後才決定 S1 寫什麼 |
| 3 | 追求量不追求質 | 每篇必過 3-Model QG ≥ 8.0 |
| Agent | Advances Leader O? | Unique Value? |
|---|---|---|
| 0. Leader | — (IS the O) | Guardian — accepts/rejects all deliverables based on whether they create genuine conviction, not just process advancement |
| 1. Prospect Researcher | YES — without specific knowledge of the prospect's content gap, the demo cannot show "your business transformed"; generic demos never create the "not a toy" feeling the O requires | YES — only agent doing pre-call intelligence gathering; creates the personalized context that all downstream agents depend on for O alignment |
| 2. Demo Builder | YES — the demo moment IS the core O delivery: if the prospect sees their own business in the demo, they believe "this is a tool for me" immediately | YES — only agent creating the live demo experience with prospect-specific article generation; Outreach Writer uses the dossier in text, Demo Builder materializes it as working product output |
| 3. Outreach Writer | YES — outreach that leads with a specific content gap insight creates the first "this company understands" feeling that pre-conditions the prospect for the "not a toy" demo revelation | YES — only agent responsible for cold-to-engaged conversion via written communication; Demo Builder shows the product, Outreach Writer creates the desire to see it |
| 4. Onboarding Manager | YES — first article published = tangible proof that "this is a real tool"; without Day 7 activation, the demo conviction fades and converts to churn, not O | YES — only agent managing the post-conversion activation phase; handles install + first article production + the "aha moment" timing that O's "I can't believe we weren't doing this before" depends on |
| 5. Success Tracker | YES — even the best demo and onboarding fail if the prospect goes cold unnoticed; Success Tracker ensures O is delivered to 100% of prospects who reach install, not just those who stay loud | YES — only agent monitoring pipeline health across all active prospects simultaneously; creates the discipline infrastructure that makes O repeatable at scale |
將整個 team 資料夾複製到你的 .claude/skills/ 目錄，即可在 Claude Code 中使用。