Data Quality Audit

Use this skill when the runtime data itself may be wrong, stale, incomplete, or inconsistent.

Trigger Conditions

Use this skill when:

• a dataset, table, or event stream has suspect values
• nullability, uniqueness, foreign-key, or enum integrity may be violated
• freshness or completeness is in question
• duplicate, drifted, or orphaned records are suspected
• the team wants a data-trust pass before a release or downstream use

Do not use this skill for migration planning, code-path debugging, or schema design.

Workflow

1. Identify the dataset under review and its purpose.
2. List the relevant quality dimensions: correctness, completeness, freshness, uniqueness, referential integrity, schema conformance.
3. Surface anomalies and likely failure modes.
4. Separate critical issues from low-priority cleanup.
5. Recommend remediation and verification.
6. Note follow-up monitoring or contracts to add.

Output

For non-trivial work, provide:

• dataset under review
• quality dimensions checked
• key anomalies and severity
• recommended remediation
• verification or monitoring suggestions

Coordination

After data audit:

• use data-migration-safety-review if remediation needs schema or backfill changes
• use systematic-debugging if anomalies look code-driven
• use docs-writer if data contracts or runbooks need to be published

Invocation Examples

• "Use data-quality-audit from dev-workflow-kit to assess freshness, completeness, and correctness of this dataset."
• "Use data-quality-audit to check whether this table has duplicate or orphaned rows before we expose it."
• "Use data-quality-audit because downstream reports look wrong but the code path seems fine."