lead-triage

bởi anthropic

Chấm điểm các khách hàng tiềm năng đến từ HubSpot dựa trên tín hiệu tương tác, mức độ phù hợp của công ty và dấu hiệu khẩn cấp để tạo danh sách "gọi 5 khách này hôm nay" kèm điểm trao đổi, soạn thảo…

npx skills add https://github.com/anthropics/knowledge-work-plugins --skill lead-triage

Lead Triage

Quick start

Pull inbound leads from HubSpot, score them, and surface a ranked call list with talking points. Drafts follow-ups and proposes calendar slots — never sends or books without owner approval.

User: "prioritize my leads"
→ Pull contacts: lifecycle stage Lead or MQL, status ≠ Unqualified
→ Score each across engagement, company fit, urgency, recency
→ Return ranked list (size adapts to volume) with talking points
→ Offer to draft follow-ups and propose calendar slots

Workflow

  1. Pull leads from HubSpot. Fetch contacts with lifecyclestage = Lead or MQL and hs_lead_status ≠ Unqualified. Use the field list in reference/hubspot-scoring.md. If HubSpot is unavailable — not installed, not connected, or erroring — do not stop. Say so plainly and run the "No HubSpot?" path at the bottom of this file: ask for a CSV export or a rough pasted list, and score that instead.

    • Empty portal check. If the pull returns zero contacts of any lifecycle stage, the portal is empty, not filtered wrong. Say so in one line, offer HubSpot's own setup (manage_onboarding, action SET_GOAL) or the "No HubSpot?" path, and do not diagnose lifecycle stages on a portal with nothing in it. The count is the test: get_user_details returns an onboarded flag, but a portal can report onboarded: false while holding hundreds of contacts, so the flag alone never decides this.
  2. Clarify if trigger is ambiguous. If the user said only "pipeline" without a qualifier, ask: "Quick pipeline overview (deal stages + total value) or prioritized call list?" — then route accordingly. Do not score leads on a bare "pipeline."

  3. Score each lead. Apply the four-dimension model in reference/hubspot-scoring.md:

    • Engagement — email replies, opens, site visits in HubSpot (last 30 days only)
    • Company fit — industry and employee count vs. owner's ICP (default: any industry, 1–50 employees). Read these from the company associated with the lead, not from the contact record — see the scoring reference. When leads have no associated company and a lead-data connector (Apollo or Clay) is connected, the enrichment leg in the scoring reference fills industry and size from the lead's email domain. It is metered — say how many rows it would enrich and what that costs in credits, and run it only on the owner's yes for this run. Nothing enriched is written back to the CRM from here.
    • Urgency — lead age, stage duration, notes containing "urgent / ASAP / deadline / budget approved". The keyword check requires actually fetching note bodies via the contact → notes association — follow the "Fetch note bodies" section of the scoring reference; the note count alone cannot fire it.
    • Recency penalty — subtract points if last activity was <24 hours ago (already touched today)

    Then check whether the scores actually separated. If the highest and lowest composite differ by less than 10 points, the CRM has no signal to rank on. Do not present the order as a ranking — run the flat-score path in the scoring reference, say plainly that the signals are empty, and order by CRM facts instead. A false ranking sends the owner to call the wrong person.

  4. Build the ranked list. Sort descending by composite score. Adapt list size to volume:

    • ≤10 leads → show all
    • 11–30 leads → show top 5
    • 30 leads → show top 8

    For each lead: name, company, score, one-paragraph talking point, last activity summary. If engagement signals are all >30 days old, flag: "Engagement signals are stale — approach as cold outreach."

  5. Offer follow-up drafts. Ask: "Draft follow-ups for any of these?" If yes, write one email per selected lead in the owner's voice per the shared voice profile. If that file has no profile yet, follow its "When there is no sample" instruction — ask for three emails the owner was happy with; if they decline, draft plain and neutral and say the drafts are un-voiced. Never invent a personality. Show drafts; do not send. If Mail is unavailable, hand the drafts over in chat as subject plus body for the owner to paste — that is a complete outcome.

  6. Render the triage board as an artifact. After the short chat answer — never instead of it — build an HTML page using the house artifact style (../../shared/artifact-style.md): a ranked table with each lead's composite score in mono, an urgency status pill per row (good / warn / critical), and a talking-points line under each lead. If the flat-score path fired, say so on the page and drop the pill colors to neutral.

  7. Offer calendar slots. Ask: "Propose call slots for any of these?" If yes, check Calendar for open 30-minute windows in the next two business days (avoid slots with existing events ±15 min). Propose two options per lead. Never create calendar events without explicit owner approval. If Calendar is unavailable, ask the owner for two or three windows they can offer and use those instead of inventing times.

Approval gates

  • Never send an email. Draft only; owner sends from their inbox.
  • Never create calendar events until approved by the user.
  • Never change lifecycle stage or mark a lead Unqualified unless the owner explicitly asks.
  • Never include Customer or Evangelist lifecycle contacts in the lead list.
  • If zero leads match the filter, explain why and offer to check what lifecycle stages are in use — do not fabricate a list.

Reference

Closing offer

End with one line on what the triage produced, then the most relevant next step with its trigger phrase — usually "call list" (/call-list) to add calendar blocks around the ranking. Up to two others when they fit: "write this outreach" (outreach-composer) or "update the CRM" (crm-autopilot). Never more than three, and never re-offer something the owner already declined this session.

No HubSpot?

Paste or upload the lead list — a CSV export or even a rough list works. Scoring and talking points run the same; the ranked call list comes back as chat plus a file instead of writing to the CRM.

Using a tool that isn't listed

The connectors named in this skill are the tested paths, not a wall. If the owner wants this flow to use a tool that isn't connected or listed, offer build-connector — it checks the connector directory first and connects through Zapier otherwise, never hand-building against a raw API. Once the connection exists, the tool joins this skill like any other optional connector, under the same approval gates.

Thêm skills từ anthropic

analyzing-financial-statements
anthropic
Kỹ năng này tính toán các tỷ số và chỉ số tài chính chính từ dữ liệu báo cáo tài chính để phân tích đầu tư.
applying-brand-guidelines
anthropic
Kỹ năng này áp dụng nhận diện thương hiệu và phong cách doanh nghiệp nhất quán cho tất cả tài liệu được tạo ra, bao gồm màu sắc, phông chữ, bố cục và thông điệp.
creating-financial-models
anthropic
Kỹ năng này cung cấp bộ công cụ lập mô hình tài chính nâng cao với phân tích DCF, kiểm tra độ nhạy, mô phỏng Monte Carlo và lập kế hoạch kịch bản cho đầu tư…
board-minutes
anthropic
Soạn thảo biên bản cuộc họp hội đồng hoặc ủy ban theo định dạng của bạn. Tự động phát hiện các cuộc họp hội đồng và ủy ban sắp tới từ lịch của bạn, yêu cầu chương trình họp và…
crm-cleanup
anthropic
Quét HubSpot để tìm các deal đã cũ, danh bạ trùng lặp và các trường bị thiếu, sau đó sửa những gì chủ sở hữu phê duyệt. Chấp nhận đối số phạm vi tùy chọn cho deals, contacts,…
redshift-api
anthropic
Chạy SQL trên Amazon Redshift — gửi câu lệnh, kiểm tra trạng thái, phân trang kết quả và duyệt cơ sở dữ liệu/lược đồ/bảng. Sử dụng công cụ này bất cứ khi nào người dùng muốn…
ticket-deflector
anthropic
Đọc email hoặc ticket của khách hàng được chuyển tiếp, lấy trạng thái đơn hàng/hoàn tiền từ PayPal và lịch sử tài khoản từ HubSpot, soạn thảo phản hồi phù hợp với giọng điệu của chủ sở hữu…
reg-feed-watcher
anthropic
Kiểm tra các nguồn cấp dữ liệu quy định ngay bây giờ và báo cáo những gì mới kể từ lần kiểm tra trước, được lọc theo ngưỡng trọng yếu của bạn. Sử dụng khi người dùng nói "kiểm tra các nguồn cấp dữ liệu",…