菜单
Unified code review: --local (diff review), --pr (adversarial PR review), --apply (batch fixes from last review).
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
Batch-verify and delete dead code: rg-confirm zero callers for N symbols, delete files, run the build once, stage results.
Run benchmarks and check regressions, add benchmark gating to CI, or audit GitHub Actions workflows. Determines mode from user intent.
Batch-apply findings from the most recent --local review. One confirmation, one test run.
Orchestrated code review of git diff using parallel agents. Standard mode (diff-scoped) or full mode (adds Roslyn semantic analysis). Produces severity-grouped plan.
Audit .NET solution health across 13 dimensions: dependency graph, solution sync, CPM compliance, package health, framework consistency, IVT & encapsulation, build config, analyzer config, NuGet config, suppression hygiene, output type & AOT, public API surface, and lock files & pruning. Works with any .NET solution (.slnx or .sln). Parses all config files once, spawns 3 parallel agents, and produces a severity-grouped report with a Mermaid dependency graph. Triggers on: solution audit, audit solution, project graph, dependency audit, solution health, audit projects, check solution.
Scans source and .claude/ files for anti-pattern violations, applies the top mechanical fixes, and reports pre/post counts.
| name | review |
| description | Unified code review: --local (diff review), --pr (adversarial PR review), --apply (batch fixes from last review). |
| when_to_use | code review, review code, audit code, full review, deep review, comprehensive review, review pr, review branch, pre-merge review, adversarial review, review this branch, check before merge, code review pr, final review, fix all findings, apply review, apply all fixes, fix code review issues. |
| argument-hint | [--local | --pr | --apply] [args] |
| user-invocable | true |
You are a dispatch router. Detect the review mode and hand off to the focused sub-skill.
Check $ARGUMENTS and the triggering message for mode keywords:
| Mode | Keywords | Dispatches to |
|---|---|---|
--local | review change, review code, audit code, full review, recent changes, last N days/commits, --perf, --test-quality | /review-local |
--pr | review pr, review branch, pre-merge, adversarial, check before merge, --pr [branch] | /review-local --full (deep pre-merge review) |
--apply | fix all, apply review, apply findings, fix all findings | /review-apply |
Default: --local if no clear PR or apply keyword. --pr takes precedence when a branch/PR number is mentioned; it runs the local review in full mode for maximum coverage before merge.
Once mode is determined, invoke the target sub-skill with the remaining arguments:
Skill("review-local", "{remaining args}")
Skill("review-local", "--full {remaining args}")
Skill("review-apply", "{remaining args}")
Keep all review logic in the dispatch targets — this router only dispatches.