data-quality-audit

name	data-quality-audit
description	Comprehensive data quality assessment against business rules, schema constraints, and freshness expectations. Activate when validating production pipelines, onboarding new data sources, or when stakeholders report data quality concerns.

When to use

• A production pipeline needs validation before a business-critical report runs
• Stakeholders report suspicious numbers or missing values
• Onboarding a new data source that needs a baseline quality profile
• A scheduled data quality SLA check is due

Process

1. Completeness analysis — detect nulls across all columns; calculate per-column and overall completeness scores
2. Duplicate identification — find exact and near-duplicate rows; distinguish intentional versioning from pipeline errors
3. Referential integrity validation — measure orphan rates for foreign key relationships documented in the schema
4. Range validation — test numeric and date fields against documented business rules (e.g. age 0–120, dates not in the future)
5. Freshness assessment — compare data currency against pipeline SLAs; flag stale tables
6. Severity classification — map findings to quality dimensions (completeness, validity, consistency, timeliness) and assign severity (critical / major / minor)

Inputs the skill needs

• Required: target dataset (path, DataFrame, or DB connection + table name)
• Required: schema documentation identifying key fields and relationships
• Required: business rules defining valid ranges, required fields, and acceptable values
• Optional: error rate thresholds (defaults: >5% nulls on required field = critical; >1% duplicates = major)
• Optional: pipeline schedule and SLA for freshness check

Output

• HTML audit report with per-dimension scores and finding details
• Markdown quality scorecard summarising dimensional performance