data-storage

Think Through Data & Storage

Overview

Production guidance for data modeling and storage. Each reference paper captures the correctness work that schema-first drafts miss: precision hazards, constraint-race conditions, index design from real predicates, soft-delete interactions with uniqueness, file-handling abuse, derived-store rebuilds, and retention obligations.

Core principle: Data outlives code. Every column, identifier, file, and derived index is a durable decision with invariants, an owner, and a lifecycle — not a convenient shape for today's feature.

Domain Law

NO DATA OR STORAGE CHANGE WITHOUT:
1. the minimum required primary paper(s) for the data being modeled selected from the context table;
2. the paper's pre-change questions
   answered, or each open point labeled as an assumption;
3. "Existing-codebase checks" run when changing an existing schema or store;
4. every applicable MUST mapped to a decision (constraint, index, policy),
   a test, or a documented exception — never silently downgraded;
5. whenever money is in scope — even a storage-only field — an explicit money
   contract names exact representation, currency/scale, each rounding or
   no-rounding boundary and mode, and deterministic boundary tests.

When to Use

Use this skill when thinking through, reviewing, changing, or verifying:

• tables, documents, relationships, denormalization, and metadata modeling;
• identifier generation, exposure, and migration (auto-increment, UUIDv4/v7, ULID, snowflake);
• timestamps, timezones, DST, scheduling, and clock-skew handling;
• money and numeric precision, rounding modes, and serialization fidelity;
• primary/foreign/unique/check constraints and constraint race conditions;
• indexes (composite, partial, covering) and query plans, N+1 elimination;
• soft delete, restore, purging, and unique-constraint interaction;
• retention, archival, expiration, anonymization, and legal hold;
• immutable/append-only records, correction records, and tamper evidence;
• file upload/download, MIME and signature validation, signed URLs, orphan cleanup;
• image/video processing pipelines and original preservation;
• full-text/vector search and index synchronization;
• record versioning, provenance, source-of-truth boundaries, reconciliation, and cleanup jobs.