Skip to main content
Back to the field guide

A field guide to the /flux-health skill

AI Data Quality and Pipeline Health Check

Data goes stale silently. /flux-health checks freshness, schema drift, null rates, orphaned records, and pipeline status with a prioritized fix list.

Flux · Data7 min readFebruary 7, 2026

Data goes stale silently. The pipeline that runs nightly fails on a Tuesday and nobody notices until an analyst pulls Friday's numbers and realizes the data is three days old. The schema in staging diverges from production because a migration applied in one and not the other. Null rates climb on a column that used to be reliable because an upstream change quietly stopped populating it. Orphaned records accumulate because a foreign key was loosened during a hotfix and never tightened back. Each issue is a quiet decay; none of them announces itself.

The /flux-health skill audits the data layer across five dimensions: freshness against expected SLA, schema drift between environments, null rates on supposedly-non-null fields, orphaned records from missing FK constraints, and pipeline run status. The output is a ranked findings list so the team fixes the data quality issue users will notice first.

What the audit covers

Freshness: each table compared against its declared SLA (e.g. invoice table should have rows from <2h ago). Schema drift: column count and types compared between staging and production. Null rates: each not-null column checked for actual nulls (sometimes a slow migration left them). Orphaned records: foreign key candidates checked for orphans. Pipeline status: scheduled jobs checked for recent successful runs.

How /flux-health works

The skill connects to the project's databases and orchestrator (Airflow, Dagster, Prefect) and runs the five checks. Findings are ranked by user-visible impact: a freshness issue on the customer dashboard is higher priority than schema drift in a backfill table. Each finding has a remediation step and the proposed long-term fix (e.g. add a freshness monitor, tighten the FK constraint with migration).

The single most useful long-term fix is freshness monitoring as alerts. /flux-health surfaces the gap and recommends the alert configuration so the team catches future freshness issues before users do.

Tonone's /flux-health skill audits data freshness, schema drift, null rates, orphaned records, and pipeline status with ranked remediation.

Install

/flux-health ships with the Flux agent in Tonone for Claude Code. Install Tonone, configure database access, and the skill audits the data layer.

1. Add to marketplace

$ claude plugin marketplace add tonone-ai/tonone

2. Install Flux

$ claude plugin install flux@tonone-ai

Frequently asked questions

What does /flux-health do?
It audits the data layer across freshness, schema drift, null rates, orphaned records, and pipeline run status, with ranked remediation.
How do I install /flux-health?
Install Tonone for Claude Code via tonone.ai/get-started. /flux-health ships with the Flux agent.

Pairs well with