renewal-predictor — Claude Code skill: predict customer churn risk
The renewal-predictor skill analyzes seven account health dimensions—engagement, support, adoption, satisfaction, billing, stakeholder continuity.
MIT 249 installs ★ 440 updated 2026-10-20
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a claude-codeWhat renewal-predictor does
renewal-predictor ingests account data from CSV, JSON, YAML, XLSX, CRM exports, or communication logs. It discovers files in the working directory, extracts metrics across seven dimensions (Engagement Frequency 20%, Feature Adoption 20%, Support Ticket Volume 15%, Usage Trends 15%, and 10% each for others), normalizes each to a 0–100 scale, and calculates a composite Health Score. Classification follows: 80–100 Likely to Renew, 60–79 Neutral/Monitor, 40–59 At Risk, 0–39 Likely to Churn.
For each account, the skill detects churn and expansion signals, identifies root causes, and selects the highest-impact intervention from its save-plays library—for example, schedule a QBR with the account champion by a specific date, introduce a new reporting module, or conduct a financial risk review. Priorities are set by ARR at risk, renewal proximity, and signal severity. Output goes to renewal-forecast.md with evidence cited for every recommendation—exactly which data points produced it.
The skill handles incomplete data gracefully. Missing dimensions are flagged with a confidence penalty; templates are requested where needed. Recent signals outweigh older ones. Across a portfolio, it identifies systemic themes (four accounts lost their champion this quarter) and surfaces them in an executive summary.
What it helps with
- forecast which customers are at risk of non-renewal
- calculate account health score across seven dimensions
- surface compound churn signals from engagement, support, and usage data
- generate a prioritized action plan with clear owner and deadline
- identify upsell opportunities in healthy accounts
Example prompts for the agent
- Predict renewal risk for all accounts in the CRM export inside data/
- Assess account health for ACME across engagement, support tickets, and usage trends over the last quarter
- Pick a save play and build an action plan for the three accounts renewing within 90 days
How to install renewal-predictor
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a claude-code2Manually: put the files in place
~/.claude/skills/renewal-predictor/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/renewal-predictor/
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a codex2Manually: put the files in place
.agents/skills/renewal-predictor/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a cursor2Manually: put the files in place
.cursor/skills/renewal-predictor/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install OneWave-AI/claude-skills renewal-predictor2Manually: put the files in place
.github/skills/renewal-predictor/SKILL.md1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a gemini-cli2Platform CLI
gemini skills install https://github.com/OneWave-AI/claude-skills --path renewal-predictorManually: put the files in place
.devin/skills/renewal-predictor/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about renewal-predictor
How do I install renewal-predictor in Claude Code?
Run npx skills add OneWave-AI/claude-skills --skill renewal-predictor -a claude-code, then open ~/.claude/skills/renewal-predictor/SKILL.md. On other platforms, the flag changes: Cursor uses -a cursor, GitHub Copilot uses gh skill install OneWave-AI/claude-skills renewal-predictor.
What does renewal-predictor do?
It computes a Health Score for each account (0–100) from seven dimensions: engagement, support, adoption, satisfaction, billing, stakeholder continuity, and usage. From this score, it predicts renewal likelihood, detects churn signals, and generates specific interventions—with owner, deadline, and success metric.
What data does renewal-predictor need?
Any format works: CSV, JSON, YAML, XLSX, CRM exports, meeting notes, or logs. If data is missing, the skill requests the appropriate templates. Incomplete data does not stop analysis—it flags which dimensions lack information and adjusts confidence accordingly.
Source text
SKILL.md · MIT
--- name: renewal-predictor description: Predicts client renewal likelihood based on health score signals. Analyzes engagement, support, adoption, satisfaction, billing, stakeholder, and usage data to calculate a weighted Health Score (0-100), classify renewal risk, and generate a prioritized action plan with early warning signals and recommended save plays. tools: Read, Write, Bash, Grep, Glob model: inherit --- # Renewal Predictor Predict client renewal likelihood by computing a multi-dimensional Health Score and generating an actionable, evidence-backed risk assessment for every account. ## Core Mission For every account, answer three questions: 1. Will this client renew, and with what confidence? 2. What signals are driving that prediction? 3. What specific intervention should the team execute right now? ## Contents - `references/dimension-rubrics.md` -- 0-100 scoring criteria, data sources, red/green flags for all 7 dimensions - `references/confidence-and-scoring.md` -- composite formula, thresholds, confidence calibration, missing-data rules, edge cases - `references/signals.md` -- compound churn and expansion signals with severity tiers - `references/save-plays.md` -- intervention templates by priority - `references/output-template.md` -- exact structure for `renewal-forecast.md` - `references/data-templates.md` -- CSV templates to request when data is missing ## Health Score Model The Health Score is a composite metric from 0 to 100, built from seven weighted dimensions. Score each dimension independently on a 0-100 scale using `references/dimension-rubrics.md`, then combine with these weights: | Dimension | Weight | |---|---| | Engagement Frequency | 20% | | Support Ticket Volume and Sentiment | 15% | | Feature Adoption | 20% | | NPS/CSAT Scores | 10% | | Billing History | 10% | | Stakeholder Continuity | 10% | | Usage Trends | 15% | Map the composite score to a prediction category (80-100 Likely to Renew, 60-79 Neutral/Monitor, 40-59 At Risk, 0-39 Likely to Churn). See `references/confidence-and-scoring.md` for the formula, thresholds, and confidence rules. ## Execution Protocol 1. **Discover available data.** Use Glob to find CSV, JSON, YAML, or XLSX files, CRM exports, meeting notes, or communication logs in the working directory. Use Grep for account names, metric patterns, and keywords like "churn," "cancel," "renew," "escalat," "competitor," "discount," "downgrade." If no data is found, request the templates in `references/data-templates.md` and stop. 2. **Parse and normalize.** For each account, extract values for as many of the seven dimensions as the data supports. Normalize every metric to the 0-100 scale in `references/dimension-rubrics.md`. Flag insufficient, missing, or ambiguous dimensions. Record the raw evidence behind each score. 3. **Compute Health Scores.** Score each dimension, apply the weights, compute the composite, map to a prediction category, and determine confidence per `references/confidence-and-scoring.md` (including the missing-data rules). 4. **Detect signals.** Scan for churn and expansion signals in `references/signals.md`. Cross-reference across dimensions for compound signals. Tag each with severity and triggering evidence. 5. **Generate interventions.** For At Risk or Likely to Churn accounts, identify the root cause, select the most impactful play from `references/save-plays.md`, and assign priority by ARR at risk, renewal proximity, and signal severity, with a clear owner, deadline, and success metric. For Likely to Renew accounts with expansion signals, suggest a specific upsell or cross-sell motion and its trigger. 6. **Write the forecast.** Generate `renewal-forecast.md` following `references/output-template.md`. Support every claim with evidence; make every recommendation actionable and specific. ## Behavioral Guidelines 1. Evidence over intuition. Every score must cite the specific data that produced it. If data is missing, say so explicitly. Never infer a score without evidence. 2. Conservative predictions. When signals conflict, weight negative signals more heavily. A false At Risk flag is cheaper than a missed churn signal. 3. Actionable specificity. "Improve engagement" is not a recommendation. "Schedule a QBR with [Champion Name] by [Date] to review Q3 usage trends and introduce the new reporting module" is. 4. No false precision. Report whole-number scores. Do not claim 92% confidence when data covers three of seven dimensions. 5. Temporal awareness. Weight recent signals more heavily than older ones, and note the recency of the data behind each score. 6. Portfolio thinking. Identify systemic themes (e.g., "4 of 12 accounts have champion turnover this quarter") and surface them in the executive summary. 7. Renewal proximity urgency. Accounts renewing within 90 days get elevated priority regardless of health score. 8. Think in ARR. Frame every classification and priority in terms of revenue impact. 9. Assume nothing about data maturity. Adapt depth and precision to the available data, and state what additional data would improve the forecast. 10. Never refuse to produce a forecast because data is incomplete. Work with what is available, flag what is not, and deliver with clear confidence markers.
© OneWave-AI · MIT · OneWave-AI/claude-skills