← ClaudeAtlas

nl-sql-querylisted

Use when a user asks a data question in natural language and wants the SQL built, executed (read-only), and results shown — for PostgreSQL/MySQL/MS-SQL via the nl-sql MCP. Explore the schema with tools, then author and run a dialect-appropriate SELECT and present the results.
ch4570/vulpora · ★ 1 · API & Backend · score 67
Install: claude install-skill ch4570/vulpora
# nl-sql-query — natural language → SQL query workflow A repeatable workflow that takes the user's **natural-language data question**, explores the DB schema with tools, authors and runs a **read-only SELECT** matched to the target dialect, and then presents the results as a table. DB access is handled by the companion MCP server `nl-sql`. This skill is a guide for **how to drive that MCP**, and does not reimplement DB connection or query execution. > **Authority**: the truth of the physical schema is the live DB, and `describe_table`/`search_objects` > are the single source for it. Do **not** guess the schema — confirm it with tools before > authoring. Dialect rules are in > [`reference/kb/dialect-differences.md`](reference/kb/dialect-differences.md); the deeper > rationale is in [`reference/principles.md`](reference/principles.md) · > [`reference/kb/`](reference/kb/INDEX.md). ## Invocation triggers - **Natural-language data questions** such as "show me the counts by order status", "how many members signed up last month?", "the 10 posts with the most comments". - "Build this question into SQL, run it, and show me the results." - When the target DB is one of PostgreSQL / MySQL / MS-SQL and the `nl-sql` MCP is connected. ## Prerequisite: the `nl-sql` MCP must be installed and configured - The `nl-sql` MCP connects to **a single DB**, configured with one `dialect` ∈ {postgres, mysql, mssql} plus connection info (read-only). - Exposed tools: `list_schemas`, `list_tables`(sch