dbaudit

Installation
SKILL.md

🔎 dbaudit

Give a balanced assessment: what is working well, what is deficient, and what could improve. A score of 9.0 or more counts as “good”, but do not call a database unconditionally good when the workflow gate finds open Critical/High findings, unused/copy status, or more than 10 % of its objects carrying unresolved needs-data flags. State the score-based result and those risk conditions separately, in the gate block (🚦 The gate).

Where this fits

For a redesign, continue with dbdesign, then dbdiagram and dbpropose when a complete recommendation is wanted. Optionally, harness/dbworkflow.py coordinates these stages; it exists only in a clone of the skills repo, and this skill never needs it. The 🔬 flags this audit raises — hot paths, real index usage, plan behaviour, cardinality — are settled by dbquery when plans and workload statistics are available.

This skill evaluates and stops. It does not recommend a redesign, write target DDL or decide what to drop — an audit that also prescribes has an incentive to shape its findings toward the prescription. Evidence, theory and a number; the decisions belong to the next skill.

Installs
2
First Seen
Oct 26, 2026