Adversarial Swift Gate

A Swift language review gate — pass/fail: a single blind reviewer subagent judges the work against this gate's rules with an adversarial mandate, and the work passes only when every rule is PASS or N/A. This skill renders verdicts; it never fixes the work.

The rules are filtered down to the checks a reviewer can decide from code evidence alone. The gate judges the failure modes agents and experienced-but-not-expert developers actually produce: continuations that hang, loops that ignore cancellation, serialized independent awaits, hand-synchronized derived state, discarded underlying errors, hand-maintained case lists that rot, boolean parameter rows, O(n²) collection accumulation, and index arithmetic that corrupts on slices.

When to Apply

• A Swift diff, feature, or PR is about to merge and needs an objective PASS/FAIL, not advisory feedback — any target: app logic, server-side Swift, CLI tools, packages.
• An agent (Claude, Codex) authored the code and you want an independent check that it is not reproducing the classic concurrency and error-handling failure shapes (unresumed continuation branches, try? await Task.sleep posing as a cancellation check, catch blocks that throw payload-free domain cases) or hand-rolling what the compiler synthesizes (CaseIterable, Comparable, memberwise inits, definite-initialization checking).
• Pre-concurrency or Objective-C-era Swift is being modernized and the ported surface needs auditing against current language affordances (@unknown default, rethrows, canImport, reduce(into:), count(where:), String(validating:)).

Do not apply to non-Swift codebases (the reviewer prompt's precondition aborts with "GATE NOT APPLICABLE"), or when the user wants explanations and refactors rather than a verdict. SwiftUI-specific review (state ownership, view identity, update cost, task lifecycle, lists, accessibility) is owned entirely by the sibling adversarial-swift-ui gate — this gate's rules are framework-independent and stay in scope alongside it. Rules whose remedy needs a newer toolchain than the project's are N/A, not FAIL — the reviewer prompt carries the version-gate table.

Review Protocol

Follow these steps exactly — the gate's value is that every review runs the same way.

Two preconditions before step 1 — both exist because their violation is the gate family's recorded field-failure mode (July 2026: gates run as cleanup drivers on a moving target rendered zero verdicts while the diff grew to 33 files):