DE//AGENT
Explore
Tutorial Lesson 10 of 10 · The Python & SQL Series community intermediate

Capstone — the SQL correctness checklist

Close the series by turning nine lessons into one standing checklist: a grain statement for every SELECT, a cardinality justification for every join, NULL-safe anti-joins, explicit casts, lint in CI, and one decision rule for slow queries. Each line cites the lesson that earned it, and the checklist ends where the review agent ends — with findings that cite severity, location, rule, and fix.

At a glance

When A new query, model, or PR needs one pass that covers the whole series; A quarterly review needs one page that covers the whole python-sql series
Where SQL
Why Nine lessons become production-ready when each rule on this checklist is a standing check with an owner, not a memory.

Learning objectives

  • Assemble the series' rules into one checklist you can run on any query
  • Turn each checklist line into a standing check with an owner
  • Emit findings with severity, location, rule, and fix

What you need

  • The series lessons 1-9 completed
  • [object Object]
  • [object Object]

Steps

  1. Line one, from lesson 1 — ask the first question: what is the grain of this result? If the author can't answer, stop. (from skill-sql-review )
  2. Line two, from lesson 3 — every join states its cardinality (1:1, N:1, N:N) and justifies it. (from skill-sql-review )
  3. Line three, from lessons 3 and 4 — prove a join cannot fan out before writing it. (from skill-sql-review )
  4. Line four, from lesson 8 — NOT IN where the subquery can return NULL is banned; use NOT EXISTS. (from skill-sql-review )
  5. Line five, from lesson 8 — comparisons across types must be explicit casts, not implicit. (from skill-sql-review )
  6. Line six, from lesson 5 — integrate SQLFluff into CI/CD pipelines so the structure rules hold without review time. (from rule-pack-sqlfluff-core )
  7. Line seven, from lesson 6 — for the slow query, decide between an index, a rewrite, or a stats refresh. (from skill-sql-query-plan )
  8. The standing output — a findings list (severity, location, rule, fix) and a grain statement for each SELECT. (from agent-sql-reviewer )

Check yourself

  • You can run the whole checklist on a query you have never seen and produce findings with severity, location, rule, and fix.
  • Each checklist line names the lesson and the rule that backs it.
  • The checklist is wired into CI — lint and review rules run without a human remembering them.

Worked example

The standing review output — from agent-sql-reviewer 

Findings list (severity, location, rule, fix)
Grain statement for each SELECT

Install the review skill the checklist runs on — from skill-sql-review 

npx de-agent-platform add sql-review

Sourced from

Security scan

Last sweep 2026-10-24: 0 finding(s), 0 marked as examples.

Common gotchas

  • A checklist that lives in a doc rots — wire the lint and review rules into CI instead.
  • Stop at the first violation; five findings at once hide which fix worked.
  • The checklist checks correctness first — the plan reading comes after the query is right.

Production considerations

  • Assign an owner to each line — an unowned check is a wish.
  • Keep the 12-case evaluation fixture in CI so the review rules stay honest.

Diagram

capstone checklist
   |
   |- grain statement --> every SELECT (lesson 1)
   |
   |- join cardinality + no fan-out --> every join (lessons 3-4)
   |
   |- NOT EXISTS + explicit casts --> every predicate (lesson 8)
   |
   |- sqlfluff in CI --> structure (lesson 5)
   |
   |- one fix per slow query --> index, rewrite, or stats refresh (lesson 6)
   |
   |- findings --> severity, location, rule, fix

Registry stats

Verification community · Oct 23, 2026
Completeness 81
Trust multiplier ×1.25
Composite 34.15 · adequate
Freshness 1 days since last source check
Difficulty intermediate
Version 0.1.0
License CC-BY-4.0
Security flags none