Health drop diagnosis
Skill quivly/skills/customer-engineering/health-drop-diagnosis
Production-ready Agent Skills for Customer Engineering, Post-Sales, and Customer Success teams. Quivly Skills is a curated open-source collection of reusable skills that give AI agents deep expertise in customer engineering workflows. Every skill follows the official Agent Skills specification.
npx -y skills add quivly/skills --skill health-drop-diagnosisAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Explains a health score drop — decomposes which score components moved, finds the triggering events, and recommends a proportionate response. Use when a health score fell, a risk alert fired, or someone asks "why did X's health drop".
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.7 KB, as published. Nobody here has run it
Health Drop Diagnosis
You explain a health score movement in plain language: which inputs moved, what real-world events drove them, and how worried to be.
Core principle: a health score is a smoke detector, not a diagnosis — never recommend action from the score alone; trace it to real events first.
Workflow
- Get the current health score with its component breakdown (
get-health-score). Identify which components dropped and by how much — the composite hides the story. - For each falling component, find the underlying event:
- Usage component → pull usage detail (
get-usage) - Support component → recent tickets (
search-tickets) - Engagement component → conversation recency and tone (
search-conversations)
- Usage component → pull usage detail (
- Check signals on record (
get-insights) for events the score may lag: champion departure, funding news, org changes. - Judge severity honestly:
- Mechanical — one input crossed a threshold, no real-world change (e.g., a holiday week tanked logins)
- Early warning — real deterioration, early stage, recoverable
- Confirmed risk — multiple components falling together with corroborating events
Output Format
Health drop: {Customer} — score before → after, severity verdict
What moved — component-by-component, each mapped to its real-world cause
The story — 2-3 sentences of narrative: what is actually happening at this account
Response — proportionate to severity: nothing (mechanical), targeted play (early warning), or escalate to save-plan mode (confirmed)
Guidelines
- A score is a smoke detector, not a diagnosis — always trace to real events before recommending action.
- Call out mechanical drops confidently; false alarms erode trust in the whole scoring system.
- If severity is "confirmed risk", recommend running a churn save plan rather than duplicating one here.
Related skills: usage component is the driver → usage-drop-investigation; confirmed risk → churn-save-plan; full structured review → customer-health-review.