跳到主内容
智客 ZICQ

技能库 智客分类:文档办公 review-team

Review Team

当用户要求由一组专业审查员代理进行代码审查,想要多个独立审查员视角,或者要求以单通或迭接固定至清洁模式运行审查员时使用. 推出重点子剂,用于正确性、安全性、结构、惯例、简洁性、UX、可靠性、遥测、测试、兼容性和文献审查.

1897 安装量

官方网址:skills.sh

技能介绍

先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。

做什么

当用户要求由一组专业审查员代理进行代码审查,想要多个独立审查员视角,或者要求以单通或迭接固定至清洁模式运行审查员时使用. 推出重点子剂,用于正确性、安全性、结构、惯例、简洁性、UX、可靠性、遥测、测试、兼容性和文献审查.

何时用

用户要求由一组专业审查员代理进行代码审查,或要求多个独立审查员视角,或要求以单通或迭接固定至清洁模式运行审查员。 为正确性、安全性、建筑性、常规、,

代理如何加载

按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Review Team、Modes、Reviewer Set、Launch Pattern、Aggregation、Iterate Mode。

文件分析

文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。

官方 description(原文)

Use when the user asks for a code review by a fleet of specialized reviewer agents, wants multiple independent reviewer perspectives, or asks to run reviewers in single-pass or iterative fix-until-clean mode. Launches focused subagents for correctness, security, architecture, conventions, simplicity, UX, reliability, telemetry, testing, compatibility, and documentation review.

Review TeamModesReviewer SetLaunch PatternAggregationIterate ModeFinal Response

来源分类:skills.sh agent-skill

SKILL.md 与 Agent 调用

官方规范 ↗
name
review-team
description
Use when the user asks for a code review by a fleet of specialized reviewer agents, wants multiple independent reviewer perspectives, or asks to run reviewers in single-pass or iterative fix-until-clean mode. Launches focused subagents for correctness, security, architecture, conventions, simplicity, UX, reliability, telemetry, testing, compatibility, and documentation review.
  1. 发现技能客户端向 Agent 提供名称与描述目录。
  2. 匹配与调用用户指定或任务匹配后,载入 SKILL.md 指令。
  3. 按需加载按步骤读取参考文档、使用脚本与素材。

具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗

安装这个技能

Skills CLI ↗

先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。

交给 Agent 安装

复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。

把 Agent Skill「review-team」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-f7ea82ebca5fa23b-%E5%AE%A1%E6%9F%A5%E5%B0%8F%E7%BB%84.html
请存为 .cursor/skills/review-team/SKILL.md 或 .claude/skills/review-team/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。

GitHub 完整包 ↗

终端安装 · Skills CLI

需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。

npx skills add 'https://github.com/cline/skills' --list

npx skills add 'https://github.com/cline/skills' --skill 'review-team'

CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。

阅读排版

name: review-team description: Use when the user asks for a code review by a fleet of specialized reviewer agents, wants multiple independent reviewer perspectives, or asks to run reviewers in single-pass or iterative fix-until-clean mode. Launches focused subagents for correctness, security, architecture, conventions, simplicity, UX, reliability, telemetry, testing, compatibility, and documentation review.

Review Team

Run a fleet of specialized reviewer subagents against the same change. Each reviewer is intentionally unaware of the others and must stay within its assigned focus area.

Reviewer prompts live in reviewers/. Read only the reviewer files you plan to launch.

Modes

  • single: Default. Launch reviewers once, aggregate findings, deduplicate, and report the final review.
  • iterate: Launch reviewers, fix accepted findings, rerun relevant reviewers, and repeat until the review is clean or the remaining findings are intentionally deferred.

If the user does not specify a mode, use single. Use iterate only when the user explicitly asks to keep fixing, make reviewers happy, or run until clean.

Reviewer Set

Core reviewers:

  • reviewers/correctness-reviewer.md
  • reviewers/security-abuse-reviewer.md
  • reviewers/architecture-reviewer.md
  • reviewers/code-quality-conventions-reviewer.md
  • reviewers/simplicity-scope-reviewer.md

Specialist reviewers:

  • reviewers/product-ux-accessibility-reviewer.md
  • reviewers/performance-reliability-reviewer.md
  • reviewers/telemetry-observability-reviewer.md
  • reviewers/testing-strategy-reviewer.md
  • reviewers/api-compatibility-reviewer.md
  • reviewers/documentation-dx-reviewer.md

Default to all reviewers when the user asks for the full team. For narrower requests, launch only the relevant reviewers.

Launch Pattern

For each reviewer subagent, provide:

  1. The exact reviewer prompt from its Markdown file.
  2. The task description, PR/issue context, or user request.
  3. The diff or changed files to review.
  4. Any test results, logs, screenshots, or relevant repository context.

Add this wrapper to every reviewer prompt:

You are one reviewer in a fleet of independent code review agents. You cannot see the other reviewers. Stay strictly within your assigned role. Return only concrete, actionable findings with file and line references. If you find no issues in your area, say so clearly. Do not provide general commentary, praise, summaries, or duplicate concerns outside your mandate.

Prefer running independent reviewers in parallel. Do not ask multiple reviewers to solve the same broad task in the same way; their reviewer prompt is their boundary.

Aggregation

After reviewers finish:

  1. Merge duplicate findings.
  2. Drop findings that lack a concrete failure mode, user impact, security risk, or maintenance cost.
  3. Resolve conflicts using the codebase, tests, and stated requirements as evidence.
  4. Rank findings by severity and practical importance.
  5. Present findings first, with file and line references.

The final review should not expose raw reviewer transcripts unless the user asks for them.

Iterate Mode

In iterate mode:

  1. Run the initial reviewer set.
  2. Decide which findings are valid and should be fixed.
  3. Implement fixes directly.
  4. Run tests or checks appropriate to the changes.
  5. Rerun only the reviewers relevant to the changed areas or unresolved findings.
  6. Continue until no actionable findings remain, or until any remaining findings are explicitly documented as out of scope, false positives, or accepted tradeoffs.

Keep the user informed between iterations. Do not loop indefinitely; stop when additional iterations are not producing material new findings.

Final Response

For single mode, return the aggregated review.

For iterate mode, return:

  • What was fixed.
  • What verification ran.
  • Which reviewers were rerun.
  • Any remaining findings or accepted tradeoffs.

相关技能

文档办公

Ontology

为结构化的代理内存和可堆肥技能所打入的知识图. 在创建/征服实体(Person, project, Task, Evention, Document)时使用,链接相关对象,强制约束,规划多步动作作为图变,或技能需要共享状态时使用. 触发到"记住","我知道什么","链接X到Y",…

文档办公

Nano Pdf

使用纳米-pdf CLI编辑带有自然语言指令的PDF.

文档办公

Word / DOCX

创建,检查,并编辑有可靠样式的Microsoft Word文档和DOCX文件,编号,跟踪更改,表格,章节,并进行相容性检查. 当 (1) 任务涉及 Word 或 ".docx " 时使用; (2) 文件包括跟踪的更改,评论,字段,表格,模板,或页面布局限制; (3) 文档必须在不…

文档办公

Api Gateway

通过 Maton 网关调用第三方 API, 它为用户已经连接的应用程序注入证书 。 当用户命名一个连接的应用程序和其中的具体行动时使用此技能——读取一个邮箱,查询一个CMM,归档一个问题,更新一个电子表格,通过一个连接的搜索或刮取提供者运行一个查询. 每一次通话都到用户连接的应用…