ADR-006 · monorepo · Ratified
ADR-006: Class A–D Data Classification & Idempotency Replay Sanitization
Prevent Class A (Secrets) and Class B (Sensitive PII) fields from ever persisting in idempotency caches or structured logs.
Context
Under Kenya’s Data Protection Act (2019) and GDPR, caching full API responses for idempotency or logging payloads can inadvertently leak National IDs, KRA PINs, or phone numbers into Redis and log aggregators.
Decision
Classify all schema fields into Class A (Restricted), Class B (Sensitive PII), Class C (Operational UUIDs/Enums), and Class D (Public). Only Class C and D fields may enter idempotency replay stores or NDJSON logs.
Consequences
Provable compliance with ODPC regulations and zero PII exposure in telemetry pipelines.