The Review skill
A one-shot audit of a task's staged artifacts, right before you deploy. Type /review and the AI checks what it built against what you asked for, field-level security, deploy order, and anything that could touch existing records — always at the Smarter AI level.
What Review does
Skills are a family — /help/planning is the flagship, and Review is the second one. Where a planning session runs ahead of the build, asking questions, Review runs after it: a single pass over the artifacts already staged on the task, checking them against what you actually asked for before you commit to a deploy.
It checks five things:
- Does it match the request — any spec gaps, like a missing filter or the wrong field type
- Field-level security — is every custom field covered by a permission set, and does that permission set auto-assign
- Deploy order — if the staged artifacts depend on each other, is the order correct
- Undeclared dependencies — anything a flow or validation rule refers to that isn't itself a staged artifact
- Safety — whether a new validation rule or flow could block saves on existing records
If it finds something, it explains the gap and suggests a concrete fix — sometimes a new artifact to save, sometimes a manual step like picking a permission set. If everything checks out, it says so plainly and leaves the deploy option in place so you can proceed.
Running it
Type /review in the composer, or press / and pick Review from the skill list — both send the same message and run the same one-shot audit. Unlike a planning session, Review never opens a multi-turn session: there's no start card, no session lane, just one turn with the findings.


What it costs
A Review turn debits credits like any Smarter AI turn — the cost is shown in the chat once it completes. There's no separate charge beyond that one turn.