churn-watch

Scores account sets for churn risk with per-account evidence links and tiered rankings.

Updated Sep 11, 2026
One-click install
npx skills add https://github.com/JYeswak/grok_bot_playground --skill churn-watch-jyeswak
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: churn-watch
Source: https://github.com/JYeswak/grok_bot_playground/tree/main/plugin/skills/churn-watch
Command: npx skills add https://github.com/JYeswak/grok_bot_playground --skill churn-watch-jyeswak

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Churn risk hides across disconnected surfaces — usage drops, support spikes, and billing friction each look benign alone. This Skill combines all three signal surfaces into one ranked, evidence-backed risk list so you know which accounts need attention first, without guessing. ## Core Features & Use Cases - Multi-surface risk scoring: Combines usage, support, and billing signals per account into low, watch, or at-risk tiers ranked by consequence. - Evidence-linked scores: Every score carries the linked signals that produced it — ticket links, invoice links, usage windows, and dates — so no score is an unverifiable rumor. - Honest unscored handling: Accounts with unreadable surfaces are marked UNSCORED with the unreachable surface named, never defaulted to healthy. - Use Case: Before a quarterly renewal review, run the skill on your book of business to get a dated markdown list of at-risk accounts with evidence links, watch-tier accounts trending wrong, and healthy counts. ## Quick Start Ask the assistant to score this quarter's account set for churn risk using usage, support, and billing signals, and produce a ranked risk list with evidence links.

Frequently Asked Questions about churn-watch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I identify which customer accounts are at risk of churning?▼

Provide the account set by stable id and the skill pulls usage, support, and billing signals per account, then combines them into at-risk, watch, and healthy tiers. Every score includes the evidence links and dates that produced it.

What signals are used to score churn risk?▼

Three surfaces are combined: usage signals (active-use trend, feature breadth, last-active date), support signals (ticket counts, severity, escalations), and billing signals (overdue status, failed charges, plan changes). No single surface promotes an account alone without corroboration.

Can this skill contact at-risk accounts or file tickets?▼

No. The skill is strictly read-only: it never contacts accounts, changes plans, issues credits, files tickets, or edits records. Any intervention belongs to a separate routine with its own approval.

What happens when an account's data cannot be read?▼

The account is marked UNSCORED with the unreachable surface named, rather than defaulted to healthy. The output records what access would settle those accounts, and stale-signal scores are flagged as stale.

Does churn risk scoring work for large low-touch account sets?▼

Yes. The output shape supports both small high-touch sets with deep per-account evidence and large low-touch sets where tier counts dominate and links are attached only to the top risks.