Tytrana

Data Model Skills Ty — DevOps skill for Claude Code

DevOps community

Agentic, read-only database assistant for SQL Server and Azure SQL.

How to install Data Model Skills Ty

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Tytrana/data-model-skills-ty and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Data Model Skills Ty does

Agentic, read-only database assistant for SQL Server and Azure SQL. Ask in plain English: the agent signs in, maps the schema into a data model, writes and runs SELECT-only queries through dbvr, and checks its answer against the results. Works with Cursor, Claude Code, and other agents.

Alternatives in DevOps

Claudebox — A Claude Code Docker Development Environment for running Claude AI's coding assistant in a fully containerized 964 ★
Check Schema Version — Check the Flyway schema version of the SQLite database on a Docker volume 205 ★
Obsidian Web MCP — Secure remote MCP server for Obsidian vaults -- access your notes from Claude, your phone, or any MCP client 99 ★

README

SQL from a data model

An agentic, read-only database assistant for SQL Server and Azure SQL. You ask a question in plain English, such as *How many orders does customer C-1001 have?*, and the agent does the database work from start to finish:

Sets itself up. It checks that dbvr is installed and every database can be reached. It opens a sign-in window for your username and password, never the chat. Then it tells you whether the login can change data.
Learns the database. It runs 7 read-only catalog scripts and writes a data model: a map of every table, column, and key.
Answers. It picks the right database and tables from the data model, writes the query, runs it, and checks its answer against the result before replying.

It never changes data. The query filter keeps it read-only whichever login you use; a read-only login is still the safest choice. The script refuses anything but `SELECT` and `WITH`. The agent doesn't guess either: it uses only tables and columns in the data model, and only values that came back from the database.

It is one skill, `/sql-from-data-model`, with two options:

Option A: Automatic. The agent does all three steps above and answers in the chat. Needs dbvr.
Option B: Manual. The agent writes the query. You run it in DBeaver and paste the result back. Use this when you can't install dbvr.

This skill replaces the earlier `/sql-from-data-model-manual` and `/discover-data-model` skills. The manual skill is now Option B. The discover steps are now part of the skill: Option A runs them for you, and Option B walks you through them when you don't have a data model yet.

dbvr uses the `microsoft` driver and a SQL username and password. The skill does not use Azure AD. It does not support other database engines.

This repo does not include a data model. The default location is the `model_path` from `paths`, inside the settings folder, which is outside this repo. You can choose another path. Keep the data model ou