build

For data and analytics engineers. In Claude Code this is a /truequery slash command. In Muse Code, describe the task in your own words, or load the skill with /skill truequery. In Codex the skill isn't here yet, only the checks.

/truequery build a customer lifetime value mart

It follows the procedure from start to finish, in your stack. You get code, tests that ran, a contract and a receipt.

model, ingest and orchestrate also work: they start build with that intent.

What it asks#

It works out whether this is a load, a transform or a schedule, then asks only what it can't find out, each with a default:

• A load: full, incremental or CDC; the unique key; the late-data window; re-runs; schema drift.
• A transform: the grain, the keys, and the incremental key and lookback.
• A schedule: schedule and time zone, catch-up and backfill, retries, alerting.

A real build#

On the demo shop, three of the four questions it asked:

1. "What kind of lifetime value should the mart hold?" Realized, to date (Recommended)
2. "Should subscription revenue (MRR) count toward LTV, or only order revenue?" Orders only, MRR shown apart (Recommended)
3. "Which customers get a row?" All customers, flagged (Recommended)

From the result:

• One row per customer: 600 rows, matching stg_customers exactly.
• Test and deleted customers: they keep their rows but show zero lifetime value. Use is_ltv_population to filter them out for headline figures.

When a request conflicts#

If you say "just append the feed, keys don't matter" and a contract says otherwise, it tells you before writing, and follows the contract unless you confirm. It never rebuilds raw data unasked.