Today's Agent Skill: Proactive Customer Churn Risk Alert
What It Does
By the time a customer asks to cancel or simply goes quiet, they have usually already made up their mind. This skill spots the early warning signs — falling purchase frequency, shrinking order size, unresolved support tickets — before customers disappear, giving you a window to act while there is still time.
How It Works
The agent analyzes your customer purchase history and support ticket data on a weekly cadence, flags any account whose behavior has meaningfully changed from their personal baseline, and drafts a personalized outreach message for each at-risk customer. It scores each risk as high, medium, or low so you know exactly where to focus your attention first.
How to Deploy It
Connect the skill to your point-of-sale, e-commerce platform (Shopify, WooCommerce, or Square), or CRM via CSV export or API in Claude Code, Gemini CLI, Cursor, or Windsurf. Run it every Monday morning. High-risk accounts go into an immediate personal outreach queue; medium-risk customers get a check-in email drafted and ready to send.
SKILL.md — Ready to Deploy
## Description
Analyzes customer purchase and support history weekly, flags accounts showing early churn signals, scores each by risk severity, and drafts personalized outreach messages for at-risk customers.
## Trigger
Run on a weekly Monday schedule, or when the user says "identify at-risk customers," "churn report," "who hasn't ordered lately," or "flag inactive accounts."
## Input
- Customer transaction CSV: customer_id, name, email, order_date, order_value, product_category
- Support ticket summary (optional): customer_id, ticket_date, status, sentiment (positive/neutral/negative)
- Engagement thresholds (user-defined): e.g., "flag customers who haven't ordered in 45 days when their average is 21 days"
- Optional: CRM notes or tags identifying VIP or contract customers
## Steps
1. Calculate each customer's baseline purchase frequency and average order value from the last 6 months of history.
2. Flag customers whose time-since-last-purchase exceeds 1.5x their personal baseline.
3. Additionally flag customers whose last 3 order values have declined more than 30% compared to their 6-month average.
4. Cross-reference with support tickets: customers with an unresolved or negativCopy the full SKILL.md and drop it into your agent's skills directory to activate this skill.