dbalisted

Postgres-first DBA workflows — run a full database health audit, design and implement schema (tables, constraints, RLS, indexes, migrations + tests), clean up unused indexes / bloat / dead tuples, diagnose and fix slow queries, and investigate live database incidents (high CPU, lock storms, error spikes, connection exhaustion). Use whenever the user says "audit the database", "DB health check", "design a schema", "model this table", "add a table for X", "clean up indexes", "database bloat", "why is this query slow", "optimize this query", "explain analyze this", "investigate the database", "the DB is on fire", "high CPU on Postgres", "connection pool exhausted", or types "/dba" — even when they don't say the word "DBA" or name a specific workflow. Discovers each project's migration, test, and deploy conventions at runtime rather than assuming them.
mvdmakesthings/skills · ★ 0 · AI & Automation · score 65

Install: claude install-skill mvdmakesthings/skills

# DBA — Postgres database administration playbooks You are running an end-to-end DBA workflow: a health audit, a schema design + implementation, a cleanup, a slow-query fix, or a live-incident investigation. This skill teaches the *how* — the sequence, the safety discipline, and the artifacts each workflow must produce. It deliberately does **not** hardcode any one project's conventions: every run starts by discovering how *this* project does migrations, tests, and deploys, and then works inside those conventions. Two ideas govern everything below: 1. **Databases outlive applications.** App code gets rewritten; schemas and data persist for years. Optimize for clarity, invariants, and recoverability over cleverness. 2. **The database is usually production.** A careless `EXPLAIN ANALYZE` on a DELETE, an index build that blocks writes, a kill command that re-ignites — DBA mistakes are immediate and user-visible. The safety model below is not bureaucracy; it is the job. ## Rules vs. playbooks — delegation This skill owns *workflows*: how to run an audit start to finish, how to investigate an incident, in what order, with what gates. It does not own rule-level Postgres knowledge (which index type to pick, data-type selection, connection-pool sizing, lock-mode tables, EXPLAIN-node interpretation). For rule-level detail, check whether the `supabase-postgres-best-practices` skill is available in this session. If it is, defer to it and **cite the rule you applied** (e.g. "per `q