review-gauntletlisted

# Review Gauntlet > **Judging doctrine:** the producer-grader split, evidence-on-every-verdict, > deterministic-aggregation, and primary-owner dedup rules this profile runs on > live in [`_shared/judge-core.md`](../_shared/judge-core.md). review-gauntlet > does not re-implement them — it inherits them through review-deep's engine > (review-deep is one of judge-core's reference implementations, §10). review-gauntlet is a **thin profile over [`review-deep`](../review-deep/SKILL.md)** — not a separate review engine. It runs review-deep's **code lenses** (the `--reviewers code` set) with review-deep's deterministic aggregation, but in a **LEAN configuration**: positional args instead of named flags, and a terse PASS / NEEDS-WORK report instead of review-deep's JSON audit-trail sidecar. It exists so the common case — "review this diff, fast, give me a verdict" — has a one-line invocation with no sidecar bookkeeping, while the heavy case (high-stakes substrate / schema / key-shape diffs that want the full audit trail, plan-conformance, runtime lenses, or `--prior-sidecar` persistent- disagreement tracking) reaches for `review-deep` directly. **review-gauntlet does NOT define its own reviewer prompts.** The lens definitions, severity rubric, anti-pattern catalog, evidence discipline, and aggregation rules are review-deep's — review-gauntlet points at them so the two never drift as a duplicate pair. If you need to change how a lens reasons, edit review-deep; the change flows thro