PostgreSQL
Your agent now designs and reviews PostgreSQL schemas against production-grade rules — proper data types (TIMESTAMPTZ, NUMERIC, JSONB), indexing for actual access paths, constraints, partitioning, and workload-aware patterns (insert-heavy / update-heavy / upsert).
What it installs
| Artifact | Path (in your project) | Purpose |
|---|---|---|
| Skill | .agents/skills/postgresql-table-design/ | Protocol the agent follows when designing tables, migrations, or queries — sourced from wshobson/agents |
| Rule | .claude/rules/postgresql.md | Always-loaded pointer: triggers on any PostgreSQL schema, migration, or query work |
How it works
PostgreSQL schema mistakes compound silently — TIMESTAMP without timezone, FLOAT for money, missing indexes on foreign keys, premature denormalization. By the time you notice, you're hauling terabytes through bad plans.
The skill encodes hard rules (use TIMESTAMPTZ for event time, NUMERIC for money, BIGINT IDENTITY for PKs, index FK columns manually), normalization guidance (3NF first, denormalize only with measured ROI), and workload-specific patterns for insert-heavy, update-heavy, and upsert paths.
The rule loads a pointer into every session, so the agent consults the skill before writing CREATE TABLE, a migration, or a non-trivial query — no explicit invocation needed.
Setup
No env vars or external accounts required. The skill is fetched from GitHub during harness-kit add via npx skills add.
Pairs well with
supabase— pairs RLS + auth patterns on top of these schema fundamentalsspec-driven— design schemas from contracts before writing DDLcode-review-gates— enforce data-type and indexing rules on every migration