本文へ移動
cccskills
無料GitHub で公開

onboarding-milestones

Daily runbook for tracking new customer accounts against onboarding milestones in Postgres and HubSpot for {{projectName}}. Covers the milestone framework, stall and overdue detection, nudge drafting, the owning-CSM alert, and the human approval gate before anything sends or changes account state.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md7.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

<skill name="onboarding-milestones"> <overview> A new account's onboarding progress only matters if someone acts on a stall while it's still cheap to fix. A daily sweep reads HubSpot for every account inside its onboarding window, checks Postgres for that account's activation milestones and product events, and classifies it as on track, stalled at a specific milestone, or overdue for full activation. Every stalled or overdue account gets a nudge email drafted to its specific missed step, and its owning CSM gets an alert with the exact context. Nothing reaches the customer and nothing on the account changes until a human reviews it.

Fresh session per sweep — state lives on the HubSpot account record itself (a kortix_onboarding_last_nudge_milestone property marks the last milestone nudged), not in a local ledger file. A sweep can find several accounts at once; handle each as an independent unit — a failure checking one account's milestones is logged and skipped, never blocking any other account in the same sweep. </overview>

<when-to-load> - The daily onboarding sweep fires. - A human asks the agent to check a specific account's onboarding status or re-run the sweep. - The milestone framework (what counts as each step, expected timing, nudge tone) changes and needs to be reflected in the next run. </when-to-load> <workflow>

Step 0 — Orient: find accounts inside the onboarding window

Query HubSpot for accounts (company or deal, whichever carries {{customer_since_property}}) where that date falls within the last {{onboarding_window_days}} days:

GET /crm/v3/objects/companies/search
  filterGroups: [
    { propertyName: "{{customer_since_property}}", operator: "GTE", value: <today - {{onboarding_window_days}}d> }
  ]
  properties: ["name", "hubspot_owner_id", "{{customer_since_property}}",
               "kortix_onboarding_last_nudge_milestone", "kortix_onboarding_last_nudge_at"]

For each match, resolve hubspot_owner_id to the owning CSM's name and note how many days into onboarding the account currently is.

Step 1 — Pull milestones and events from Postgres (read-only)

SELECT milestone, completed_at
FROM activation_milestones
WHERE account_id = :account_id
ORDER BY completed_at;

SELECT event_name, occurred_at
FROM product_events
WHERE account_id = :account_id
  AND occurred_at >= now() - interval '{{onboarding_window_days}} days'
ORDER BY occurred_at;

Adapt table and column names to what the project's schema actually calls them; the shape above is illustrative. The milestone rows are the source of truth for what's done — product events are supporting evidence for why a milestone hasn't completed (e.g. zero logins vs. logins with no key action).

Step 2 — Classify against the milestone framework

MilestoneWhat it meansExpected by
Workspace liveFirst login after signupDay 1
Core setupKey integration or configuration completedDay 3
First activation eventThe core "aha" action completed — the first time the account got real valueDay 7
Team expansionA second teammate invited and activeDay 14
Full activationRecurring usage across most of the last 14 daysDay 30 (or {{onboarding_window_days}})

For each account:

  • On track — every milestone whose expected day has passed is completed.
  • Stalled at <milestone> — the account is past that milestone's expected day without completing it, but still inside the onboarding window.
  • Overdue — past {{onboarding_window_days}} days total without full activation, regardless of which earlier milestones did complete.

An account can only be stalled at its earliest incomplete milestone — don't report team expansion as the stall if core setup itself never finished.

Step 3 — Draft the nudge (stalled or overdue accounts only)

Skip this step if kortix_onboarding_last_nudge_milestone already equals the milestone this account is currently stalled at — it was nudged for this exact stall on a prior run. Draft again only when the stalled milestone has changed (the account moved past the previously-nudged one and is now stuck further along, or regressed to an earlier open item) or the marker is missing.

Write the nudge addressed to the customer's specific missed step — not a generic "how's it going" check-in:

  • Name the one concrete action tied to the stalled milestone (e.g. "connect your first data source" for a core-setup stall, not "finish setup").
  • Keep it short, helpful, and low-pressure — an offer to help, not a warning.
  • Reference what the account has already done, if anything, so it doesn't read as a form letter to someone who's made partial progress.

Hold the draft in {{draft_channel}} for the owning CSM to review, edit, and send.

Step 4 — Alert the owning CSM

Post to {{alert_channel}}, addressed to the owning CSM, with:

  • The account name and how many days into onboarding it is.
  • Which milestones are completed and which is the current stall (or that it's overdue).
  • The nudge that was drafted (or, if this exact stall was already nudged, a note that no new draft was made and the prior one is still pending).

This alert is not deduplicated — it reappears every day the account remains stalled or overdue. It's a working reminder for the CSM, not a one-time notice.

Step 5 — Write the marker back

For every account that got a new nudge drafted in Step 3, set kortix_onboarding_last_nudge_milestone to that milestone's name and kortix_onboarding_last_nudge_at to now. This is the only HubSpot write — never the deal stage, the owner, a plan, or a seat count.

</workflow> <guardrails> - **Draft and alert only.** The nudge email is always a draft in `{{draft_channel}}` for the owning CSM to review, edit, and send. The agent never emails the customer directly. - **No account changes, ever.** The agent never changes a plan, a seat count, a billing setting, the deal stage, or the account owner — in HubSpot or any other system. - **HubSpot writes are scoped** to `kortix_onboarding_last_nudge_milestone` and its timestamp. Nothing else on the record is touched. - **No duplicate nudges for the same stall.** Always check `kortix_onboarding_last_nudge_milestone` before drafting — an account stalled at the same milestone as last run isn't re-nudged, though a change in which milestone it's stalled at always gets a fresh draft. - **CSM alerts aren't deduplicated.** Unlike the nudge draft, the Slack alert reappears every day the account remains stalled or overdue — it's meant to be seen until someone acts. - **Scoped secrets.** Postgres and HubSpot access is brokered server-side through connectors; no raw credential is ever pasted into chat. </guardrails> </skill>

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Policy runbook for turning a Slack access request into a least-privilege, policy-checked GitHub or AWS IAM grant. Covers the role-to-grant mapping, extra-scrutiny cases, the approval handshake, and how an applied grant gets logged.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Get a complete picture of any company or person before outreach. This skill always works with web search, and gets significantly better with enrichment and CRM data.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Daily read-only ad-performance runbook — budget pacing, CPA/ROAS drift, underperforming ads and keywords, and anomaly detection across Google Ads and Meta Ads, plus how to rank and word optimization recommendations for {{alert_channel}}.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Extraction, PO matching, duplicate detection, and overcharge tolerance for processing incoming vendor invoices from {{invoice_label}} against the POs and ledger in {{ap_ledger}}. Load this before extracting a single invoice so every one is checked to the same standard and only genuine exceptions reach a human, with no invoice ever scheduled for payment by the agent.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Daily brand-mention monitoring loop for {{brand_terms}}. Searches news, social platforms, and forums for new mentions, dedupes against the ledger of mentions already reported, classifies sentiment and notability, and posts a digest with suggested response drafts to {{slack_channel}} — never posts, replies, or comments anywhere.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

kortix-ai のスキルをすべて見る

このスキルの問題を報告する