A Spec Kit preset that bakes database conventions into the spec-driven workflow. ORM-agnostic — the rules are about SQL schema design, not a particular framework.
It appends to core templates and one command (your existing spec-kit templates and other presets keep working):
| Target | What is added | Purpose |
|---|---|---|
speckit.constitution (command) |
Database + Data protection confirmation step | Forces confirming the target DB (engine, version, migration tool) and the security posture (roles, RLS) before the constitution is drafted |
constitution-template |
Database + Data protection sections | Records engine / version / migration tool and the project-wide security posture |
spec-template |
Data Requirements | What data the feature needs — entities, reads/writes, retention, consistency |
plan-template |
Database Design | Naming conventions, OLTP/OLAP classification, normalization (3NF for OLTP) & denormalization trade-offs, schema changes, constraints & integrity (DB-enforced invariants, NOT NULL, FK actions, CHECK, soft delete × UNIQUE, case-insensitive uniqueness), transactions & concurrency (locking strategy, retries, outbox, idempotency, timeouts), data protection (column classification, secrets, RLS, least-privilege), external identifier exposure, SQL portability (ANSI-first, vendor-feature flagging), query safety (SQL injection — bound params, allowlisted identifiers), SQL correctness (NULL semantics, join fan-out, deterministic order, half-open ranges, upsert over check-then-insert), indexing plan (selectivity, index-friendly predicates), N+1 / query-volume plan, performance diagnosis (baseline → waits → plan → statistics → re-measure) |
tasks-template |
Schema Review Checklist | Run before merging any migration — naming, integrity, transactions & concurrency, SQL correctness, normalization, performance, portability, query safety, migration safety, data handling |
It also ships engine-specific reference files, not appended to any
template — the plan's "Target database" line points at the matching one so it
gets read during /speckit.plan instead of bloating every project's plan with
guidance for engines it doesn't use:
| File | Covers |
|---|---|
references/oracle.md |
Bind variables & the shared pool, NUMBER/VARCHAR2 semantics, IDENTITY vs SEQUENCE, bitmap indexes, when a PL/SQL package earns its place, materialized views |
references/mysql.md |
InnoDB vs MyISAM, utf8mb4, PK choice vs InnoDB clustering, the pre-8.0.16 CHECK gap, online DDL, gap locks, replication format |
references/postgres.md |
text/timestamptz/numeric/identity choices, exclusion constraints, un-indexed FKs, GIN/GiST/BRIN/partial indexes, CONCURRENTLY & lock-queue-safe DDL, MVCC/vacuum & autovacuum tuning, jsonb, RLS, pooling, version gates, schemas & partitioning |
references/sqlserver.md |
Clustered index vs PK, covering/filtered indexes, READ_COMMITTED_SNAPSHOT, system-versioned temporal tables, parameter sniffing |
references/dynamodb.md |
NoSQL — access-pattern-first design, single-table design, partition/sort key design, GSI/LSI, item limits, ConditionExpression/TransactWriteItems, TTL, Streams. Says plainly which generic sections don't apply. |
references/mongodb.md |
NoSQL — embed vs. reference, document validation ($jsonSchema), index design (compound/multikey/text), operator injection, multi-document transactions, TTL indexes, change streams, sharding. Says plainly which generic sections don't apply. |
On any other engine (SQLite, CockroachDB, …) there's no reference
file to read — the generic plan-template sections are the default, not a
fallback for the engines that lack dedicated guidance.
specify preset add --from https://github.com/gdakuzak/spec-kit-preset-db-standards/archive/refs/heads/main.zipUpdate an already-installed copy — add only appends, so remove first (one line):
specify preset remove db-standards && specify preset add --from https://github.com/gdakuzak/spec-kit-preset-db-standards/archive/refs/heads/main.zipInstall a specific released version instead of main:
specify preset add --from https://github.com/gdakuzak/spec-kit-preset-db-standards/archive/refs/tags/v0.8.2.zipLocal development:
specify preset add --dev ./spec-kit-preset-db-standards
specify preset resolve constitution-template # verify the section appears
specify preset remove db-standards # when done- Any project with a relational database where schema quality matters.
- Teams that want naming, indexing, and N+1 checks to be part of planning and review rather than caught in production.
- No database at all.
- A NoSQL store with no reference file yet (see
references/) — the generic templates assume a relational engine; DynamoDB and MongoDB are the covered exceptions. - You already have a house preset covering the same templates with
replace— stackingappendon top still works, but check the combined output.
Found a gap, a wrong default, or want another engine covered? Open an issue.
MIT