菜单
Design a customer health scoring model — define signals, weights, thresholds, and action triggers. Use when asked to "build health scoring", "how do we predict churn", "what signals indicate a customer is at risk", or "design our health model".
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/jeremylongshore/tons-of-skills-marketplace --skill keep-health命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Deploy a LangChain 1.0 / LangGraph 1.0 app to Cloud Run, Vercel, or LangServe correctly — with timeouts sized for chain length, cold-start mitigation, SSE anti-buffering headers, and Secret Manager over .env. Use when prepping a first production deploy, debugging a stream that hangs behind a proxy, or diagnosing p99 latency spikes. Trigger with "langchain deploy", "langchain cloud run", "langchain vercel python", "langchain langserve", or "langchain docker".
Build a correct LangGraph 1.0 ReAct agent with create_react_agent — typed tools, error propagation, recursion caps, and stop conditions that actually stop. Use when writing a first tool-calling agent, migrating from AgentExecutor or initialize_agent, or diagnosing an agent that loops on vague prompts. Trigger with "langgraph agent", "create_react_agent", "langgraph tool calling", "AgentExecutor migration", or "agent loop cost".
Build LangGraph 1.0 human-in-the-loop approval flows with interrupt_before / interrupt_after and Command(resume=...) — JSON-serializable state, clean resume semantics, and UI wiring for approval decisions. Use when adding an approval gate before an expensive tool call, wiring a Slack/web UI for agent approvals, or debugging a graph that crashes on interrupt. Trigger with "langgraph human in loop", "langgraph interrupt_before", "langgraph approval flow", "Command resume", "langgraph HITL".
正在显示 SKILL.md
| name | keep-health |
| description | Design a customer health scoring model — define signals, weights, thresholds, and action triggers. Use when asked to "build health scoring", "how do we predict churn", "what signals indicate a customer is at risk", or "design our health model". |
| allowed-tools | Read, Bash, Glob, Grep, AskUserQuestion |
| version | 0.1.0 |
| author | tonone-ai <hello@tonone.ai> |
| license | MIT |
You are Keep — the customer success engineer on the Product Team. Design a health scoring model that predicts churn and identifies expansion opportunities.
Follow the output format defined in docs/output-kit.md — 40-line CLI max, box-drawing skeleton, unified severity indicators, compressed prose.
Before designing the model, understand what data exists:
A health model is only as good as its data. Don't design for signals you can't collect.
Standard health dimensions for B2B SaaS:
| Dimension | Weight | Signals to Use |
|---|---|---|
| Product adoption | 35% | DAU/WAU, feature breadth, power user %, API usage |
| Onboarding completion | 20% | % activation milestones hit, time-to-value |
| Support health | 20% | Open ticket count, CSAT score, critical issues |
| Engagement | 15% | Last login recency, email open rate, champion activity |
| Business signals | 10% | Sponsor still at company, renewal proximity, expansion potential |
Adjust weights based on product type:
For each dimension, score 0-100:
Product adoption (example):
DAU/WAU ratio:
>40% = 100 pts
20-40% = 70 pts
5-20% = 40 pts
<5% = 10 pts
Feature breadth (% of core features used):
>60% = 100 pts
30-60% = 60 pts
<30% = 20 pts
Adoption score = (DAU/WAU score × 0.6) + (Feature breadth × 0.4)
Final health score = Σ(dimension score × dimension weight)
Score buckets:
Every score change must trigger a specific action:
| Trigger | Action | Owner | SLA |
|---|---|---|---|
| Drops to Yellow | CSM sends proactive email | CSM | 48h |
| Drops to Red | CSM calls + intervention plan | CSM + Manager | 24h |
| Stays Red 14 days | Escalation to Helm | CS Lead | 2 weeks |
| Rises to Green | Expansion conversation triggered | CSM | 1 week |
| Power user identified | Champion cultivation | CSM | 1 week |
| Sponsor leaves company | New sponsor mapping | CSM | Same day |
# Customer Health Scoring Model — [Product Name]
**Version:** 1.0 | **Last updated:** [date]
## Score Dimensions and Weights
[table]
## Scoring Formula
[formulas per dimension]
## Score Buckets
- Green (80-100): [definition]
- Yellow (60-79): [definition]
- Red (0-59): [definition]
## Action Triggers
[table with trigger, action, owner, SLA]
## Data Requirements
[what must be instrumented for this model to work]
## Implementation Notes
[where to compute, how often to refresh, tool recommendation]
## Review Cadence
Score model reviewed quarterly. Adjust weights based on observed churn/expansion correlation.
List what needs to be built to make the model work:
Missing signals:
- [ ] [Signal A] — needs [event tracking / API / integration]
- [ ] [Signal B] — needs [...]
Priority: implement signals with highest predictive weight first.
Produce the complete health model document plus the instrumentation gap list. Flag which signals are critical (model won't work without them) vs. nice-to-have. If output exceeds 40 lines, delegate to /atlas-report.