Score any at-risk account 0 to 100 and get the exact re-engagement play to run, including the move the moment a champion quits.
Churn Prevention takes one account, its usage numbers, its NPS trend, its support tickets, whether the champion just left, and turns that into a 0 to 100 health score split across product usage, engagement, and relationship health. Below 50 it lands the account in the Red tier and hands you a step by step critical intervention play. It also runs a dedicated champion departure protocol, since a champion leaving is the single biggest churn predictor the skill tracks.
The video script says it "watches usage drops, engagement, and champion departures, scores the risk, and triggers a re-engagement play." That holds up. We tested it on an invented account (fictional company, fictional churn signals, no real customer data) with a 42% usage drop, a champion who'd just left, and a dropping NPS score. It came back with a real 19/100 score, correctly placed it in the Red tier, and produced the matching intervention play and champion departure protocol, both traceable line by line to the skill's own tables. One honest caveat on the word "triggers": this is a markdown skill you run inside Claude Code, not a live monitoring system. It doesn't watch your CRM in the background and fire alerts on its own. You feed it the signals for an account, and it does the scoring and playbook writing for you, on demand.
The repo's own install scripts gave us real trouble though. Its recommended install.sh path silently installs nothing on a machine without the gh skill extension, and its alternative cc-gtm.py script crashes outright on Python 3.9 or 3.11, the versions many Macs still ship with by default. Neither is a flaw in the skill's actual content. The fix is a one line manual copy, covered below.
Anyone doing customer success or account management without a dedicated CS platform: a founder running post-sale relationships themselves, a solo CSM, or a small team that doesn't have Gainsight or Vitally. It replaces the job of manually noticing an account has gone quiet and figuring out what to do about it, and gives you a defensible, structured answer instead of a gut call.
git installed (to clone the repo)cc-gtm.py installer (not required, see Install)This downloads the full gtm-skills repository, which contains 211 skills including Churn Prevention.
git clone https://github.com/LeadMagic/gtm-skills.gitThis copies only the one skill folder into your own project's Claude Code skills directory. It's fully self-contained, everything it needs is already inside this one folder.
mkdir -p .claude/skills
cp -R gtm-skills/skills/growth/churn-prevention .claude/skills/churn-preventionThis makes Claude Code aware the skill exists so it can use it when you ask about churn or retention.
claudeThen ask it: "what skills are available in .claude/skills?" It should list churn-prevention.
Sample input used in testing (invented account, not a real company): a SaaS vendor called Ledgerloop, customer Bramwell and Voss Freight, $38,000 a year, renewal in 45 days. Usage down 42% over 30 days, no login for 16 days, NPS dropped from 8 to 5, the VP of Finance who championed the deal left 12 days ago with no replacement named yet, and there's an unresolved support ticket.
Abbreviated real output:
Product Usage: 8/40
Engagement: 6/30
Relationship Health: 5/30
Total score: 19/100 -> Red tier (<50)
Re-engagement play: Red, Critical Intervention
1. CSM calls within 24 hours, not email
2. Name the specific signal directly: "We noticed usage
dropped about 40% this month, how can we help?"
3. Build a 30-day success plan
4. Request an executive sponsor call, renewal is in 45 days
5. Weekly check-ins until the score clears yellow
Champion departure protocol triggered: congratulate on
LinkedIn, ask for an intro to the replacement, escalate to
the executive sponsor if no replacement is named in 2 weeksinstall.sh runs but reports installed=0, failed=1. Your GitHub CLI doesn't have the gh skill extension installed, which is what the script relies on. Skip it and use the manual copy in Step 2 above instead, it's simpler and doesn't depend on gh at all.
cc-gtm.py crashes with SyntaxError: f-string expression part cannot include a backslash. That script needs Python 3.12 or newer. Check your version with python3 --version. If you're on an older Python, use the manual copy in Step 2 instead, it has no version requirement.
The bundled check-output.py validator says FAIL, missing terms: retention on a deliverable that's clearly complete. That script is a simple keyword check, it's just looking for the literal word "retention" somewhere in the text, not judging quality. Add the word once anywhere in your output and it'll pass. Don't take a fail here as a sign your actual churn analysis is wrong.
Built by LeadMagic (https://github.com/LeadMagic/gtm-skills), MIT license, 53 stars at test time. Two corrections to the repo's own docs: install.sh claims it falls back to a dependency-free copy path when GitHub CLI's skill installer isn't available, but in testing it only falls back when the gh binary itself is missing entirely, not when the gh skill subcommand errors, so it can report 0 installed with no working fallback. Separately, scripts/cc-gtm.py requires Python 3.12 or newer to even parse, which isn't stated anywhere in the docs, and it hard-crashes on Python 3.9 and 3.11.
This is one agent. The 7-Day AI Native Challenge is how you learn to build the next ten yourself, live, with Dany.
Join the Challenge →Too busy to build it yourself? Focentra AI will build, install and train your team on it. Priced on the call.
Talk to Focentra AIThis guide includes affiliate links for tools mentioned in the setup steps. If you sign up through them, I may earn a commission at no extra cost to you.