check

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Output style (plain words, no dashes, no hyphens)

输出风格（通俗易懂，无破折号、连字符）

Write everything this skill produces, files and messages alike, in plain simple language. Keep technical terms that carry real meaning; explain each in plain words. Never use a dash or a hyphen as punctuation: no em dash, no en dash, and no hyphenated compounds. Write

read only

, not

read-only

. Say it in simple words, or reword the sentence. Code, file paths, command flags, and values other skills match on keep their hyphens. Use short sentences, commas, or parentheses. Clear beats clever.

本技能生成的所有内容（包括文件和消息）均采用简洁直白的语言撰写。保留具有实际意义的技术术语，并用直白语言逐一解释。绝不使用破折号或连字符作为标点：不使用长破折号、短破折号，也不使用连字符构成的复合词。请写

read only

，而非

read-only

。用简单的语言表达，或改写句子。代码、文件路径、命令标志以及其他技能需匹配的值可保留连字符。使用短句、逗号或括号表达。清晰胜于花哨。

What this skill does

本技能的功能

/check

is the gate before merge. It confirms a change is sound in two different ways, run as two modes; they are separate jobs and you usually run both, verify first:

verify
(runtime proof): run the real app and watch the change behave. Proves the feature actually works and conforms to the spec (every acceptance criterion met, every specced surface built), which green tests never reveal. Read only on code, owns no durable files. Runs on the main thread. Typically after
```
/develop
```
.
review
(fresh model code review): a rigorous senior read of the diff, on a different model than wrote the code, because a model reviewing its own output shares its blind spots. Writes findings ranked by severity to
```
docs/reviews/
```
. Read only on code. Typically before opening a PR.

Neither mode edits your code.

verify

points failures at

/debug

/develop

;

review

reports findings for the implementer to fix.

/check

是合并前的把关环节。它通过两种独立的模式确认变更是否合理可靠，通常需要依次运行两种模式，先执行 verify：

verify
（运行时验证）：启动真实应用并观察变更的运行情况。验证功能是否真正可用且符合规格要求（满足所有验收标准，完成所有规格中定义的功能模块），而这是普通测试无法体现的。仅读取代码，不生成持久化文件。在主线程运行。通常在
```
/develop
```
之后执行。
review
（全新模型代码评审）：由与编写代码的模型不同的模型对代码差异进行严格的资深评审，因为模型评审自身输出时会存在盲区。将按严重程度排序的评审结果写入
```
docs/reviews/
```
。仅读取代码。通常在发起 PR 之前执行。

两种模式均不会修改你的代码。

verify

会将问题指向

/debug

或

/develop

；

review

会将评审结果反馈给开发者进行修复。

Pick the mode (route before doing anything else)

选择模式（先确定路由再执行其他操作）

Read the argument and route:

Starts with
```
verify
```
(or
```
run
```
) → runtime proof. Read
```
modes/verify.md
```
and follow it fully. Pass any remaining arguments (a feature name, a scope) through.
Starts with
```
review
```
→ code review. Read
```
modes/review.md
```
and follow it fully. Pass the review steering through unchanged (e.g.
```
/check review with opus
```
,
```
/check review uncommitted
```
).
No mode word, or ambiguous → ask once which one, capability first (a picker where the agent has one, else plain text): question "Which check?", header "Check", options
```
verify: run the real app and prove it works (recommended right after building)
```
·
```
review: a fresh-model read of the diff before a PR
```
. The picker appends its own free text option; add it explicitly in a plain text fallback. Then read that mode's file and follow it.

Do not mix the two in one run. If the engineer clearly wants both, do

verify

first (confirm it works), then offer

review

读取参数并确定路由：

以
```
verify
```
（或
```
run
```
）开头 → 执行运行时验证。阅读
```
modes/verify.md
```
并严格遵照执行。传递所有剩余参数（如功能名称、范围）。
以
```
review
```
开头 → 执行代码评审。阅读
```
modes/review.md
```
并严格遵照执行。保持评审引导参数不变（例如
```
/check review with opus
```
、
```
/check review uncommitted
```
）。
未指定模式词或模式不明确 → 询问一次需要哪种模式，优先使用选择器（若Agent具备该功能，否则使用纯文本）：问题为“选择哪种检查？”，标题为“检查”，选项为
```
verify: 启动真实应用并验证功能可用（建议在构建后立即执行）
```
·
```
review: 在发起PR前由全新模型评审代码差异
```
。选择器会附加自由文本选项，在纯文本回退方式中需明确添加该选项。之后读取对应模式的文件并遵照执行。

请勿在一次运行中混合两种模式。若工程师明确需要同时执行两种模式，请先执行

verify

（确认功能可用），再提供

review

选项。

Portability (any OS, any agent)

可移植性（支持任意操作系统、任意Agent）

Any Agent Skills client on macOS, Linux, or Windows.

git

is the only required CLI and behaves the same everywhere; other shell snippets are POSIX reference, not literal scripts, so use your agent's own cross platform file, process, and browser tools and apply branching logic yourself. Each mode file adds its own portability notes (browser/HTTP driving for

verify

, the cross model spawn and bundled file handling for

review

). No subagent support falls back to running the work inline, noted per mode.

Bundled files live in this skill's folder:

modes/verify.md

modes/review.md

, and (for review)

review-agent-prompt.md

and

review-guide.md

. Read only the mode file you routed to; the main agent resolves the review bundled files to absolute paths when it spawns the reviewer.

支持 macOS、Linux 或 Windows 系统上的任意 Agent Skills 客户端。

git

是唯一必需的 CLI 工具，且在各系统中的行为一致；其他 shell 代码片段仅作为 POSIX 参考，并非字面脚本，因此请使用Agent自身的跨平台文件、进程和浏览器工具，并自行处理分支逻辑。每种模式的文件会添加各自的可移植性说明（

verify

模式涉及浏览器/HTTP驱动，

review

模式涉及跨模型启动和捆绑文件处理）。若不支持子Agent，则回退为内联执行任务，每种模式会单独说明。

捆绑文件位于本技能的文件夹中：

modes/verify.md

、

modes/review.md

，以及（针对review模式的）

review-agent-prompt.md

和

review-guide.md

。仅读取路由到的模式对应的文件；主Agent在启动评审模型时会将review相关的捆绑文件解析为绝对路径。