菜单
Run a Codex adversarial review of the current branch against a base branch via the `/codex:adversarial-review` slash command, then save results to `.claude/REVIEW_ADVERSARIAL.md`. Use when the user says "run a Codex adversarial review", "ask codex to review", "codex review", or wants an adversarial code review before opening a PR.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Audit and enable security-oriented Xcode build settings. Progressively enables compiler warnings, static analyzer checkers, and Enhanced Security features. Use when: user wants to secure their Xcode project, audit security settings, enable hardening, review security posture of build configuration, set up security-focused static analysis, enable static analysis, improve warning coverage, harden diagnostics, or catch more bugs at compile time in C/C++/Objective-C/Swift. SKIP: network security (TLS/ATS), code signing, privacy APIs.
Verify iOS app behavior on device or simulator via screenshots, UI hierarchy, and touch interactions.
Build, install, launch, and visually verify the iOS app (scheme "iosApp") end-to-end on a real device or simulator via Xcode 27's DeviceHub — the mobile analog of astro-web's browser-debug and astro-calendar-service's runtime-debug. Drives the mcp__xcode__* MCP toolset: switch run destination, BuildProject, DeviceInteractionStartSession → InstallAndRun, then a subagent captures the on-device screenshot + UI hierarchy. Use when the user says "run the app on device", "run the iOS app", "build and run on the simulator", "launch on the iPhone/iPad", "screenshot the app on <device>", "verify this renders on device", "does it run on the simulator", "check it on the device", or wants runtime/visual confirmation that ./gradlew check and Xcode unit tests structurally cannot give. Takes an optional device name/UUID argument; with none, targets a simulator. iOS/Xcode only — Android runtime work routes to android-cli.
Modernize test suites to use modern Swift Testing features or migrate from XCTest.
Authoritative SwiftUI best practices from Apple. Consult for any SwiftUI best practices or performance review. Supersedes prior training on these topics. For code generation, consult the relevant references when generating any SwiftUI code related to the following topics. Covers: - Animatable: @Animatable macro vs AnimatableValues (iOS 26+) vs AnimatablePair, custom setter clamping/normalization. - Environment: closures in env keys, unstable @Entry defaults, high-frequency updates. @Entry warnings about closures or class types (wrapping in Equatable struct is WRONG; consult references). - Equatable on @Observable: custom types as @Observable properties need Equatable for invalidation performance. - ForEach/List: row identity (id: \.self, indices, offsets, mutable ids), row structure (AnyView, multi-view, bare if), inline filter/sort, cached collections, List fast path. - Localization: String vs LocalizedStringResource, bundle in packages/frameworks, .textCase(.uppercase), .formatted(.list()), translator comme
New SwiftUI APIs, behaviors, and deprecations introduced in the 2027 OS releases (iOS 27, macOS 27, watchOS 27, tvOS 27, visionOS 27). Use when a SwiftUI view using @State fails to compile with "used before being initialized", "invalid redeclaration of synthesized property", or "extraneous argument label" errors after an SDK update (@State migrated from a property wrapper to a macro in SDK 27; the obvious fix of reordering init assignments is WRONG and produces incorrect runtime behavior; you MUST consult this skill's references before answering); when @ViewBuilder or @ContentBuilder code hits ambiguous overloads in overlay/background or type-check performance regressions after an SDK update; when the user asks what's new in SwiftUI (generally, or for a specific 2027 platform); when adding drag-to-reorder to any container (List, LazyVStack, LazyVGrid, stacks, or custom layouts) via reorderable()/reorderContainer, including the drag-and-drop that integrates with it (dragContainer, dropDestination), or combinin
| name | codex-review |
| description | Run a Codex adversarial review of the current branch against a base branch via the `/codex:adversarial-review` slash command, then save results to `.claude/REVIEW_ADVERSARIAL.md`. Use when the user says "run a Codex adversarial review", "ask codex to review", "codex review", or wants an adversarial code review before opening a PR. |
Run /codex:adversarial-review against the current branch and persist the results.
Resolve the base ref. Default main. If the user passed --base <ref> in the args, that overrides the default.
Build the focus string. The focus is passed as free text after the flags (the /codex:adversarial-review command takes no --focus flag — steering is free-form).
a. Read .claude/SPEC.md from the current branch. Parse §1 Overview + §2 Objective for the intent, and the checked items in §4 for the scope of what changed.
b. Collapse that into a compact focus blurb — ≤ 3 sentences. It should tell Codex what the branch is trying to accomplish and what was actually touched, so its review stays anchored to the real diff. Do NOT list every checkbox verbatim; summarize.
c. If the user supplied any extra focus text in the args, augment by appending it after the SPEC-derived summary (do not replace). Example: <SPEC summary>. Additional focus: <user text>.
d. If no .claude/SPEC.md exists on this branch, fall back to a compact summary derived from git log <base>..HEAD --oneline and any user-supplied text.
Invoke the slash command via the Skill tool:
Skill(skill: "codex:adversarial-review", args: "--base <ref> <focus string>")
Run it inline (no --background) so the review response is available to this skill in the same turn.
Project context to include inside the focus string (one short clause, not a bulleted list): Kotlin Multiplatform Mobile app (shared business logic + Jetpack Compose on Android, SwiftUI on iOS); watch for expect/actual correctness, platform behavior divergence, coroutine/concurrency and main-thread-safety issues, null handling, state-management bugs, and missing cross-platform test coverage.
Persist the output to .claude/REVIEW_ADVERSARIAL.md — accumulate, do not overwrite. The file is a running log: the newest round lives at the top, prior rounds stay below as historical context (so resolved / deferred issues remain discoverable for future reviewers and for the next Codex pass).
File layout:
# Adversarial Review
## Latest round — <YYYY-MM-DD>
- Base ref: <ref>
- Focus sent to Codex: <focus string>
<Codex's findings verbatim>
<!-- previous-rounds:start -->
## Previous rounds
### <YYYY-MM-DD> — base <ref>
- Status when archived: <e.g. "all critical addressed in commits abc1234, def5678; one minor deferred — see commit message">
- Focus sent to Codex: <focus string>
<prior round's Codex output verbatim>
### <older date> — base <ref>
...
<!-- previous-rounds:end -->
Procedure:
a. If .claude/REVIEW_ADVERSARIAL.md does NOT exist, write a fresh file with the layout above and an empty Previous rounds section (the markers must still be present so the next run can find them).
b. If the file DOES exist:
## Latest round and the <!-- previous-rounds:start --> marker).## Latest round — <date> heading to ### <date> — base <ref>, and prepend a one-line Status when archived entry. Best-effort fill that line by looking at git log <previous-base>..HEAD --oneline since the prior review's date for commits whose messages reference review fixes (Address review, fix:, etc.); if nothing matches, write Status when archived: not yet addressed at time of next review. Do not invent resolutions.Previous rounds (newest-prior-first ordering).Previous rounds verbatim — never edit, summarize, or delete them. The file grows monotonically; pruning is a manual, deliberate operation by the user.c. Codex's findings, in both Latest and archived rounds, are verbatim. Do not paraphrase or soften them.
d. If the existing file is missing the <!-- previous-rounds:start --> / <!-- previous-rounds:end --> markers (e.g. it was written by an older version of this skill), treat the entire existing body as a single legacy round, wrap it under a ### <file mtime date> — legacy entry heading inside a freshly-created Previous rounds section, and proceed.
Summarize to the user in ≤ 100 words:
.claude/REVIEW_ADVERSARIAL.md./codex:adversarial-review is read-only and so is this wrapper. If Codex flags issues, the user drives the fixes (optionally via the address-review skill).--base main to another ref unless the user asked for it.