data-privacy-governance

Classify every field as public, internal, PII, or PHI, then minimize collection to the stated purpose. Encrypt data at rest and in transit, and anonymize or pseudonymize identifiers wherever identity is not required.

Use an append-only store where each record carries event_id, UTC timestamp, authenticated actor, action, resource, outcome, and a prev_hash linking to the prior record. Hash chaining detects deletion and reordering, and retention plus legal-hold policy governs the lifecycle.

Pseudonymization is reversible with a separately stored key and still counts as personal data under GDPR. Anonymization is irreversible, typically enforced with k-anonymity plus l-diversity or t-closeness, and is preferred for analytics and exports.

Use data-quality when implementing ETL validation gates such as schema, freshness, and anomaly checks with no privacy requirement. Use data-strategy when designing enterprise catalogs, glossaries, ownership models, or lineage strategy without a privacy review.

The skill treats this as an explicit edge case requiring designed reconciliation, such as erasing personal payloads while preserving tamper-evident event metadata. Consent withdrawal must propagate erasure across copies, backups, and derived datasets.

No. It covers design and review of data flows with GDPR-centric patterns, but it is not a DPIA substitute or legal sign-off. CCPA, LGPD, and HIPAA add obligations not fully covered, and jurisdiction-specific counsel remains necessary.