SQL

Run SQL against this project's database.

How to get the connection string

Read the project's env file (e.g. .env, backend/.env) and extract the database URL (DATABASE_URL or equivalent). Never hard-code it — always read the file fresh.

Run the query

If $ARGUMENTS is non-empty, treat it as the SQL to run.

Otherwise use AskUserQuestion to ask:

• Question: "Enter the SQL to run:"
• Type: free text input

Execute with:

psql "<DATABASE_URL>" -c "<SQL>"

For multi-statement SQL (semicolon-separated), use psql "<DATABASE_URL>" <<'EOF'\n<SQL>\nEOF via bash heredoc so all statements run in one session. (Swap psql for the project's actual DB client if it isn't Postgres.)

Output

• Print the full output (rows returned, UPDATE/INSERT counts, errors).
• If the query returns rows, display them as a markdown table.
• If it fails, show the error and suggest a fix.

Notes

• Do not push any changes to git.
• For destructive queries (DROP, DELETE without WHERE, TRUNCATE) ask the user to confirm before running.
• Keep a schema reference (table → PK → key columns) in this file once the project's schema stabilizes, and update it whenever the models/migrations change — that's what makes this skill fast instead of needing a fresh schema lookup on every call.