DBA AI Skill (Super DBA)

A comprehensive, engine-aware Database Administration (DBA) toolkit designed for AI CLI agents (like Claude Code, Gemini CLI, OpenCode).

🎯 Intention & Goal

The goal of this skill is to transform your AI assistant into a Senior Database Administrator. Instead of relying on generic or hallucinated SQL advice, this skill provides the AI with strict, battle-tested rules for schema design, query optimization, indexing strategies, and zero-downtime migrations.

It is structured using an 80/20 architecture:

• SKILL.md (Root): Contains the 80% universal rules for all relational databases (SARGability, Normalization, Leftmost Prefix).
• references/ (Folder): Contains the 20% engine-specific deep dives (Execution Plans, advanced DDL gotchas, and diagnostic scripts) that the AI reads dynamically when needed.

🗄️ Supported Databases

This skill is highly specialized and handles the historical quirks, breaking changes, and performance tuning for the world’s Top 5 Relational Database Management Systems (RDBMS):

• MySQL: From 5.7 legacy traps (missing CTEs, ignored CHECK constraints) up to modern 8.0/8.4 LTS gotchas (implicit sorting removal, ONLY_FULL_GROUP_BY enforcement).
• PostgreSQL: Deep knowledge of PG 11 through 15+ breaking changes (the death of OIDs, the public schema security trap, CTE materialization changes, and B-Tree deduplication).
• Microsoft SQL Server (T-SQL): Advanced insights on DMVs, Execution Plans, and version traps (like the STRING_SPLIT ordering glitch or NOLOCK abuse).
• Oracle: Mastery of PL/SQL collections, 12c Identity columns (escaping Sequence-Trigger hell), the ROWNUM pagination nightmare, and the 23ai BOOLEAN miracle.
• SQLite: Navigating the transition from legacy “Flexible Typing” to Modern STRICT tables, the addition of DROP COLUMN support, and JSONB.

⚡ Core Capabilities

1. SQL Auditing & Review: A strict workflow for reviewing schemas and queries, classifying findings by severity (CRITICAL, WARNING, SUGGESTION, INFO).
2. Schema Design: Universal normalization rules, right-sizing data types, and engine-specific traps (like never using UUIDs as Primary Keys in InnoDB).
3. Query Optimization: Enforcing SARGability, Covering Indexes, and rewriting subqueries.
4. Execution Plan Analysis: Deciphering EXPLAIN FORMAT=JSON (MySQL), EXPLAIN (ANALYZE, BUFFERS) (Postgres), SSMS Execution Plans, and DBMS_XPLAN (Oracle).
5. Zero-Downtime Migrations: Safe DDL operations for massive production tables using the “Shadow Table” approach and online index creation.
6. Diagnostics: Health-check scripts to find unused indexes, calculate table bloat (dead tuples), and identify slow queries across engines.

🚀 Installation & Usage

To use this skill with your favorite AI CLI agent (like OpenCode, Claude Code, or Gemini CLI), you simply need to place it in the agent’s designated skills directory.