I built an AI/Copilot-oriented database investigation skill and I’m looking for feedback on whether this kind of structured workflow is useful for AI-assisted development

[miller-28]

🏷️ Discussion Type

Question

💬 Feature/Topic Area

Visual Studio

Body

Hi GitHub Community,

I recently released AI DB Investigator, an open-source skill designed to help AI approach database analysis in a more structured way.

Repository:
https://github.com/miller-28/skill-ai-db-investigator

The idea is simple:

Most AI workflows are very good at producing fast answers, but database problems usually require investigation before answers.

A slow query, missing index, unclear schema, suspicious relation, or bad data model usually cannot be solved by one direct response. It requires a sequence of technical checks.

AI DB Investigator is my attempt to give AI a more disciplined database investigation flow.

The skill is focused on helping AI reason through things like:

• schema structure
• table relationships
• indexes
• query patterns
• possible bottlenecks
• root cause separation
• safe improvement suggestions
• production-oriented database review

The larger direction is this:

AI-assisted development should not only generate code.
It should also help investigate systems.

That is especially important around databases, because database issues often sit beneath the surface. The visible problem may be a slow endpoint, but the real cause may be a missing index, a bad join pattern, weak data modeling, or growth that changed the original assumptions.

I built this as a small, focused skill rather than a large framework. The purpose is to encode a database investigation mindset into a reusable AI workflow.

I would appreciate feedback from developers, database engineers, PostgreSQL users, and people experimenting with AI-assisted engineering workflows.

Some questions I’m especially interested in:

• Would this kind of structured DB investigation skill be useful in your AI workflow?
• What database checks should be added?
• Should the skill focus deeper on PostgreSQL first?
• What real-world debugging scenarios would make good examples?
• How would you expect Copilot or another AI agent to use a skill like this?

Thanks.

[ushxyuki]

This is a really useful idea. A structured DB investigation skill would be helpful because many database issues need step-by-step checking, not just quick answers.

I think the skill should include checks for:

• missing or unused indexes
• slow queries
• table relationships
• schema design issues
• duplicate or inconsistent data
• query performance bottlenecks
• suspicious joins
• possible normalization issues

Starting with PostgreSQL first sounds like a good direction because it has strong tools like EXPLAIN and ANALYZE, and many developers use it in real projects.

Good real-world debugging examples could include:

• finding why a dashboard query is slow
• checking if a missing index is causing performance issues
• investigating duplicate records
• reviewing bad joins between tables
• identifying schema problems before adding new features

I would expect Copilot or another AI agent to use this skill like a checklist: inspect the schema, analyze queries, find likely causes, then suggest safe improvements with explanations.

Great project — this could be very helpful for developers working with AI-assisted database debugging.