Token导航 LogoToken导航

Agent Skill · 研究检索

review-implementation(审查实施情况)

review-implementation 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

review-implementation 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为操作浏览器;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 64,557。

总安装 297周安装 12GitHub Stars 175下载 93核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:review-implementation(审查实施情况)
来源仓库:https://github.com/playmoreai/agent-skills
仓库路径:skills/review-implementation
安装命令:
npx skills add https://github.com/playmoreai/agent-skills --skill review-implementation
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/playmoreai/agent-skills --skill review-implementation

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

review-implementation 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。适用于项目复盘、方案验证和实施效果评估等研究检索类任务场景。通过结构化查询支持对已有实现方案的梳理与比对分析。

安装前建议确认权限范围和维护状态,以及是否会触发联网或命令执行。可结合来源仓库和原始 README 文档核验具体用法和功能边界。

来源仓库内路径:skills/review-implementation

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/playmoreai/agent-skills --skill review-impleme…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

review-implementation 怎么安装?

在终端执行 npx skills add https://github.com/playmoreai/agent-skills --skill review-implementation。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

review-implementation 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

review-implementation 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

review-implementation 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

review-implementation 用的人多吗?

公开数据是累计安装 297 次,近一周安装 12 次,下载 93 次,GitHub 175 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

review-implementation 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
review-implementation

审查实施情况 · 本页

297175操作浏览器未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Review Implementation

Overview

Review code against markdown requirements and produce evidence-backed findings. When the user asks for fixes, apply targeted changes that bring the implementation back to the docs without expanding scope.

Modes

  • Use review-only mode when the user asks to review, assess, audit, verify, or find issues.
  • Use fix mode when the user asks to fix, update, refactor, complete, apply changes, or make the implementation match the docs.
  • In fix mode, keep changes minimal and reversible. Do not add features that are not justified by the docs.

Workflow

1) Intake and Discovery Gate

  • Read all provided.md files and extract explicit requirements, constraints, and acceptance criteria.
  • Read repo guidance such as AGENTS.md, README, architecture notes, and existing task or PRD files when relevant.
  • Inspect affected code, tests, fixtures, routes, schemas, migrations, services, components, permissions, config, and observability surfaces as needed.
  • Identify the current behavior, expected behavior, data flow, integration points, validation options, and likely blast radius.
  • Ask only when a critical decision cannot be made safely after discovery.

2) Map Requirements to Evidence

  • Build an internal traceability map: requirement -> status -> code evidence -> gap -> action.
  • Classify each requirement as satisfied, partial, missing, conflicting, overbuilt, or deferred.
  • Use file/line evidence for important claims. Do not rely on general impressions when code can be inspected.

3) Run Multi-Pass Review

Review in this order:

  1. Requirements coverage: every requirement is satisfied or explicitly unresolved.
  2. Correctness: happy paths, edge cases, errors, empty states, permissions, state transitions, and rollback behavior are handled.
  3. Integration: changed modules fit together without contract breaks, duplicated ownership, or hidden assumptions.
  4. Simplicity: the solution is no more complex than necessary.
  5. Cleanup: repeated logic, dead code, temporary code, noisy logs, unused files, and unused dependencies are removed.
  6. Security/privacy: auth, access control, secrets, sensitive data, injection risks, and audit needs are safe.
  7. Performance: expensive queries, N+1 patterns, unnecessary renders, redundant network calls, and blocking work are addressed.
  8. Validation: chosen checks are appropriate for the risk.
  9. Documentation/operability: docs, release notes, migrations, rollback, monitoring, or support notes are updated when required.

Treat overbuilt code as a review finding when it adds features, abstractions, state, dependencies, or workflow paths not required by the docs.

4) Apply Focused Fixes

  • Fix issues directly in code with minimal, targeted changes.
  • Split overly long files/functions only when the split reduces real review or maintenance risk.
  • Remove or simplify overbuilt code that is not justified by the docs.
  • Preserve existing patterns unless the docs or code evidence justify a change.
  • If PRD or phase files are part of the request, update checkboxes, validation notes, discoveries, and change logs only when implementation evidence supports the update.

5) Validate with Evidence

  • Choose the smallest sufficient validation for the risk: static checks, unit tests, integration tests, API-level E2E, browser/UI checks, simulator checks, screenshots, manual smoke checks, or observability checks.
  • Run relevant checks when available and appropriate. If checks are unavailable, too costly, or not allowed, state the gap and use the best available evidence.
  • Do not run broad or expensive validation by reflex when a narrower check proves the changed behavior.

Report Format

Lead with the result that matters most:

  • In review-only mode, list findings first in severity order with file/line references, then summarize coverage, validation, and residual risk.
  • In fix mode, summarize fixes applied, validation performed, and any remaining findings or risks.
  • If no issues are found, say so clearly and name any validation gaps.
  • For critical blockers, stop and ask. For non-critical ambiguity, make the best reasonable decision, record the assumption, and continue.

平台分布

Codex

35.96%
按下载量换算33

Claude

29.17%
按下载量换算27

Cursor

19.48%
按下载量换算18

Gemini CLI

9.65%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills