churn-signal
Skill4 starsChanged 6 months agoMITView on GitHub
Assess churn or expansion risk for a customer account or segment by pulling usage data, support sentiment, and engagement signals. Produces an account health rating and recommended intervention. Use before QBRs, renewal conversations, or proactive CS outreach.
In this repository17 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- debateSkill
- decision-briefSkill
- design-an-interfaceSkill
- docxSkill
- impact-sizingSkill
- interview-analysisSkill
- learnSkill
- parallel-prototypeSkill
- pdfSkill
- pptxSkill
- rapid-prototypeSkill
- recallSkill
- roadmap-narrativeSkill
- sprint-healthSkill
- web-browserSkill
- xlsxSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
If no input, ask:
- Distinctive: line 132
If analysing a segment rather than a single account:
- Distinctive: line 142
Don't manufacture precision: If you only have usage data and no sentiment, say the score is based on usage only.
- Distinctive: line 145
Anonymise in shared outputs: If this brief will be shared broadly, replace account names with "Account A (Enterprise, Manufacturing)" style …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Churn Signal Skill
Assess account health for a specific customer or a segment of customers. Pulls quantitative usage signals and qualitative sentiment signals, combines them into a health score, and recommends what to do.
Input: $ARGUMENTS
Accepts:
/churn-signal [account name]— Health check for a specific account/churn-signal [segment]— e.g., "enterprise accounts in manufacturing" or "trial users from last quarter"/churn-signal— Will ask interactively
If no input, ask:
"Which account or segment do you want to assess? You can name a specific customer, a plan tier, an industry, or a cohort."
Phase 1: Signal Gathering
Pull from all available sources in parallel. Never block on a failing source.
Usage Signals (Amplitude)
- MAU/WAU trend over last 90 days — growing, flat, or declining?
- Feature adoption breadth — using 1 feature or many?
- Login frequency vs 90-day average
- Any sudden drop-off events (when did it happen?)
- Power users — how many, trending up or down?
Sentiment Signals (Intercom)
- Open or unresolved support tickets
- Ticket volume trend — more or fewer issues recently?
- Sentiment in recent conversations — frustrated, neutral, satisfied?
- Any explicit churn signals ("we're evaluating alternatives", "our contract is up")
Internal Signals (Glean / Slack / Jira)
- Any CS or sales notes about this account
- Known commitments that haven't been delivered
- Competitive mentions
- Recent escalations
Contractual Context (if available)
- Renewal date (if known from any source)
- ARR value
- Plan tier
Phase 2: Score Health
Score each signal dimension on a 1-5 scale:
- 5 = Strong positive signal
- 3 = Neutral / unclear
- 1 = Strong negative signal
| Dimension | Score | Evidence |
| Usage trend | ||
| Feature adoption breadth | ||
| Support sentiment | ||
| Engagement recency | ||
| Unresolved commitments |
Overall health: Average the scores → map to:
- 4.0-5.0 → 🟢 Healthy
- 3.0-3.9 → 🟡 Monitor
- 2.0-2.9 → 🟠 At Risk
- 1.0-1.9 → 🔴 Critical
Phase 3: Output
# Account Health: [Account Name / Segment]
**Date**: [Today] | **Analyst**: AI PM Kit | **Confidence**: [High/Medium/Low]
---
## Health Score: [Score/5] — [🟢 Healthy / 🟡 Monitor / 🟠 At Risk / 🔴 Critical]
---
## Signal Summary
| Dimension | Score | Key Finding |
|---|---|---|
| Usage trend | [1-5] | [e.g., MAU down 23% in last 30 days] |
| Feature breadth | [1-5] | [e.g., Only using 1 of 4 core modules] |
| Support sentiment | [1-5] | [e.g., 3 open tickets, one flagged frustrated] |
| Engagement recency | [1-5] | [e.g., Last login 18 days ago] |
| Commitments | [1-5] | [e.g., Pending feature request from Dec] |
---
## Top Risk Factors
1. [Most concerning signal — why it matters]
2. [Second risk]
3. [Third risk, if applicable]
## Positive Signals
- [What's working for this account]
- [Any green flags that offset risk]
---
## Recommended Intervention
**Priority**: Immediate / This week / This quarter / None needed
**Recommended action**: [Specific action — not generic advice]
- [e.g., "CS to schedule a proactive call this week — reference the usage decline and ask about their Q2 plans"]
- [e.g., "PM to follow up on the outstanding feature request — it was raised 90 days ago"]
- [e.g., "Share the new Reporting module release notes — they're power users of reporting but haven't adopted the new exports"]
**Owner**: [CS / PM / Sales / Auto-communication]
---
## Data Gaps
- [Sources that returned no data]
- [Information that would sharpen this assessment]Bulk Mode (Segments)
If analysing a segment rather than a single account:
- Pull the top 10-20 accounts by ARR in the segment
- Score each one using the same framework
- Produce a ranked table from highest to lowest risk
- Highlight the top 3 that need immediate intervention
Principles
- Don't manufacture precision. If you only have usage data and no sentiment, say the score is based on usage only.
- Name the specific risk. "At risk" without explanation is useless. Say exactly what's concerning.
- Recommend something actionable. The output should tell a CS or PM what to do tomorrow, not just how bad the situation is.
- Anonymise in shared outputs. If this brief will be shared broadly, replace account names with "Account A (Enterprise, Manufacturing)" style labels.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

