Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问clear审计未展示

review_local评论本地

Agent Skill

review_local 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。

总安装

593

周安装

24

GitHub Stars

公开资料未说明

下载量

186
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

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_local(评论本地)
来源仓库:https://github.com/ikatsuba/skills
仓库路径:skills/review_local
安装命令:
npx skills add ikatsuba/skills --skill "review:local"
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。

AgentSkills.tonpx skills
npx skills add ikatsuba/skills --skill "review:local"

简介

查找、检索和筛选相关信息,适合快速定位候选结果。

  • 适用于 Codex、Claude、Cursor、Gemini CLI 等宿主环境。
  • 通过 github 安装,命令为 npx skills add ikatsuba/skills --skill "review:local"。
  • 需确认权限范围和维护状态,注意是否触发联网或文件读写。
  • review_local 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

name
review:local
description
Local Code Review - analyzes code changes and provides structured feedback before commit

Local Code Review

Performs a thorough code review of local changes before they are committed or pushed. Identifies potential issues, suggests improvements, and ensures code quality standards are met.

When to use

Use this skill when the user needs to:

  • Review changes before creating a commit
  • Get feedback on code quality and potential issues
  • Check for common mistakes and security concerns
  • Validate code against best practices

Instructions

Step 1: Determine Review Scope

Check what the user wants to review. If arguments are provided, use them. Otherwise, auto-detect:

  1. Check for staged changes first:
   git diff --cached --stat
  1. If no staged changes, check unstaged:
   git diff --stat
  1. If nothing to review, check recent commits or ask user:

- Offer to review the last commit - Offer to compare current branch with main/master - Ask user to specify what to review

Supported scopes:

  • staged - Review staged changes (default if available)
  • unstaged - Review unstaged working directory changes
  • all - Review all uncommitted changes
  • last or HEAD - Review the last commit
  • branch - Review all commits on current branch vs main
  • <file-path> - Review specific file(s)
  • <commit-hash> - Review specific commit

Step 2: Gather Changes

Based on the determined scope, gather the diff:

# Staged changes
git diff --cached

# Unstaged changes
git diff

# All uncommitted
git diff HEAD

# Last commit
git show HEAD

# Branch diff
git diff main...HEAD

Also gather context:

  • File types being modified
  • Size of changes (lines added/removed)
  • List of affected files

Step 3: Analyze Code

Review the changes for the following categories:

CategoryWhat to check
CorrectnessLogic errors, edge cases, null handling, off-by-one errors
SecurityInjection vulnerabilities, exposed secrets, unsafe operations
PerformanceN+1 queries, unnecessary iterations, memory leaks
MaintainabilityCode clarity, naming, duplication, complexity
Best PracticesLanguage idioms, framework conventions, project patterns
TestingMissing tests, test coverage for new code

For each issue found, note:

  • Severity: 🔴 Critical, 🟠 Warning, 🟡 Suggestion
  • Location: file path and line number
  • Description: what the issue is
  • Recommendation: how to fix it

Step 4: Check Project Context

Before finalizing the review:

  1. Look for existing patterns in the codebase
  2. Check for project-specific conventions (linters, style guides)
  3. Verify consistency with similar code in the project

Step 5: Generate Review Report

Create a structured summary report:

# Code Review Summary

**Scope:** [What was reviewed]
**Files:** [Number] files ([+lines added], [-lines removed])

---

## Overview

[1-2 sentence summary of the changes and overall assessment]

---

## 🔴 Critical Issues

> Issues that must be fixed before committing

### [Issue Title]
**File:** `path/to/file.ts:42`

[Description of the issue]

**Recommendation:**

[Code suggestion]


---

## 🟠 Warnings

> Issues that should be addressed

- **[file:line]** — [Brief description]
- **[file:line]** — [Brief description]

---

## 🟡 Suggestions

> Optional improvements to consider

- [Suggestion 1]
- [Suggestion 2]

---

## ✅ What Looks Good

- [Positive observation 1]
- [Positive observation 2]

---

## Checklist Before Commit

- [ ] All critical issues resolved
- [ ] Tests added/updated for new functionality
- [ ] No debug code or console.logs left
- [ ] No hardcoded values that should be configurable

Step 6: Handle Edge Cases

No changes found:

  • Inform the user
  • Suggest what they might want to review

Very large diff (>500 lines):

  • Warn user about the size
  • Offer to focus on specific files or categories
  • Prioritize critical issues

Binary files:

  • Note their presence but skip detailed review
  • Warn if unexpected binary files are included

Only deletions:

  • Verify nothing important is being removed
  • Check for orphaned references

Step 7: Offer Next Steps

After presenting the report, offer actions:

  1. Fix issues — Help fix the identified problems
  2. Re-review — Run the review again after changes
  3. Proceed anyway — If issues are acceptable, continue to commit
  4. Explain — Provide more details on any issue

Review Guidelines

Severity Levels

LevelCriteriaAction
🔴 CriticalBugs, security issues, data loss risksMust fix
🟠 WarningCode smells, potential issues, missing testsShould fix
🟡 SuggestionStyle improvements, minor optimizationsNice to have

Constructive Feedback

  1. Be specific — Point to exact lines and provide examples
  2. Explain why — Don't just say "bad", explain the consequence
  3. Suggest solutions — Provide actionable recommendations
  4. Acknowledge good code — Highlight what's done well
  5. Prioritize — Focus on what matters most

Security Checklist

Always check for:

  • Hardcoded secrets, API keys, passwords
  • SQL/NoSQL injection vulnerabilities
  • XSS vulnerabilities in user input handling
  • Insecure deserialization
  • Path traversal in file operations
  • Exposed sensitive data in logs

Arguments

  • <args> - Optional scope specification

- staged - Review staged changes - unstaged - Review working directory changes - all - Review all uncommitted changes - last - Review the last commit - branch - Review current branch vs main - path/to/file - Review specific file - abc123 - Review specific commit

Examples:

  • review:local - Auto-detect and review available changes
  • review:local staged - Review only staged changes
  • review:local src/auth/ - Review changes in auth directory
  • review:local last - Review the last commit
  • review:local branch - Review all branch changes vs main

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Claude Code

28.49%
按下载量换算53

windsurf

21.73%
按下载量换算40

OpenCode

17.07%
按下载量换算32

Codex

13.1%
按下载量换算24

Antigravity

7.51%
按下载量换算14

Gemini CLI

3.6%
按下载量换算7

安全审计

暂无安全审计结果可展示。

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

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。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills