Revue de Code Complète

VérifiéSûr

Outil d'analyse de code complet qui examine le code pour détecter les vulnérabilités de sécurité, les problèmes de performance, les violations de style, les conventions d'équipe, les modèles spécifiques au langage, l'actualité des dépendances et les meilleures pratiques. Utilisez cette compétence lors de la révision de pull requests, de fichiers spécifiques, de modifications préparées ou de l'ensemble du codebase pour garantir la qualité et la cohérence.

Spar Skills Guide Bot
DeveloppementIntermédiaire
7002/06/2026
Claude Code
#code-review#security#performance#best-practices#linting

Recommandé pour

Notre avis

Effectue une revue de code complète en vérifiant la sécurité, les performances, le style, les conventions d'équipe, les patterns idiomatiques, la fraîcheur des dépendances et les meilleures pratiques.

Points forts

  • Couvre un large éventail de catégories de revue (sécurité, performances, style, etc.)
  • S'adapte aux conventions de l'équipe en analysant les fichiers de configuration existants
  • Vérifie la fraîcheur et la sécurité des dépendances ajoutées
  • Peut analyser des fichiers, des PR, des changements staging ou l'ensemble du dépôt

Limites

  • N'execute pas le code, donc ne détecte pas les bugs runtime
  • Peut manquer des défauts spécifiques au domaine métier
  • La vérification de la fraîcheur des dépendances est limitée aux nouvelles dépendances ajoutées
Quand l'utiliser

Utilisez cette compétence pour obtenir une revue structurée et approfondie avant de merger une pull request ou après avoir effectué des modifications importantes.

Quand l'éviter

Évitez de l'utiliser pour du code non versionné ou lorsque vous avez besoin d'une exécution de tests automatisés ou d'une analyse dynamique.

Analyse de sécurité

Sûr
Score qualité88/100

The skill is a template for AI to perform code review, instructing only to read files and use git/GH commands to gather diffs. These are standard, non-destructive operations. There are no instructions for executing harmful commands, exfiltrating data, or disabling security features.

Aucun point d'attention détecté

Exemples

Review staged changes
Review staged changes in my git repository for security and performance issues.
Review a specific file
Review the file src/utils/auth.ts for vulnerabilities and best practices.
Review a pull request
Review PR #42 for style violations and team conventions.

name: code-review description: Performs comprehensive code reviews checking for security vulnerabilities, performance issues, style violations, team conventions, language patterns, dependency freshness, best practices, and code duplication. Use when reviewing code changes, PRs, or files. argument-hint: [file path, PR number, or "staged" for staged changes]

Comprehensive Code Review

Perform a thorough code review of the specified code, checking for security, performance, style, conventions, and best practices.

Input

The target for review is provided as: $ARGUMENTS

Accepted inputs:

  • File path: Review a specific file (e.g., src/utils/auth.ts) or directory containing source files
  • PR number: Review a pull request (e.g., #123 or 123)
  • "staged": Review all currently staged git changes
  • "diff": Review unstaged changes in the working directory
  • "all": All code within the repository
  • No argument: Ask the user what they want reviewed

Review Categories

Analyze the code against each of these categories and report findings:

1. Security Vulnerabilities

Check for:

  • Injection vulnerabilities (SQL, command, XSS, LDAP, etc.)
  • Authentication and authorization flaws
  • Sensitive data exposure (hardcoded secrets, credentials, API keys)
  • Insecure cryptographic practices
  • Missing input validation and sanitization
  • Insecure deserialization
  • Path traversal vulnerabilities
  • CSRF vulnerabilities
  • Improper error handling that leaks information
  • OWASP Top 10 violations

2. Performance Issues

Check for:

  • Inefficient algorithms (O(n²) when O(n) is possible, etc.)
  • N+1 query problems
  • Missing database indexes (when schema is available)
  • Unnecessary re-renders (React, Vue, etc.)
  • Memory leaks and resource cleanup issues
  • Blocking operations in async contexts
  • Excessive memory allocation
  • Missing caching opportunities
  • Inefficient data structures
  • Unoptimized loops and iterations

3. Style Violations

Check for:

  • Inconsistent naming conventions
  • Improper indentation and formatting
  • Missing or excessive whitespace
  • Line length violations
  • Inconsistent brace/bracket style
  • Import ordering issues
  • Commented-out code that should be removed
  • TODO/FIXME comments that need addressing
  • Compliance with project linter rules (if config files exist)

4. Team Conventions

Look for project-specific patterns by examining:

  • Existing code in the repository for established patterns
  • Configuration files (.eslintrc, .prettierrc, pyproject.toml, etc.)
  • Contributing guidelines (CONTRIBUTING.md)
  • Existing documentation
  • Test patterns and naming conventions
  • File and folder organization patterns
  • Error handling patterns used elsewhere in codebase

Flag deviations from established team patterns.

5. Language Canonical Patterns

Check adherence to idiomatic patterns for the source language:

  • TypeScript/JavaScript: Proper typing, async/await usage, module patterns
  • Python: PEP 8, Pythonic idioms, type hints
  • Go: Effective Go guidelines, error handling patterns
  • Rust: Ownership patterns, Result/Option usage
  • Java: Java conventions, design patterns
  • C#: .NET conventions, LINQ usage
  • (Apply similar standards for other languages)

6. Dependency Freshness

When new dependencies are added:

  • Check if the dependency is actively maintained
  • Verify the version is current (not outdated)
  • Look for known vulnerabilities in the version
  • Assess if the dependency is necessary or if native solutions exist
  • Check license compatibility
  • Evaluate the dependency's size and impact on bundle

7. Industry Best Practices

Check for:

  • SOLID principles violations
  • DRY (Don't Repeat Yourself) violations
  • Proper separation of concerns
  • Appropriate error handling and logging
  • Meaningful variable and function names
  • Functions doing one thing well
  • Appropriate abstraction levels
  • Proper use of design patterns
  • Documentation for public APIs
  • Test coverage for new functionality

8. Code Duplication (DRY)

Scan for:

  • Repeated code blocks within the reviewed files
  • Similar logic that exists elsewhere in the codebase
  • Copy-pasted code with minor variations
  • Opportunities to extract common utilities
  • Duplicated algorithms that could be consolidated
  • Repeated patterns that warrant abstraction

Output Format

Structure your review as follows:

# Code Review: [target]

## Summary
[Brief 2-3 sentence overview of the code quality and main findings]

## Critical Issues 🔴
[Security vulnerabilities and bugs that must be fixed]

## Warnings 🟡
[Performance issues, potential bugs, and significant best practice violations]

## Suggestions 🔵
[Style improvements, minor optimizations, and nice-to-haves]

## Detailed Findings

### Security
[Findings or "No issues found"]

### Performance
[Findings or "No issues found"]

### Style
[Findings or "No issues found"]

### Team Conventions
[Findings or "No issues found"]

### Language Patterns
[Findings or "No issues found"]

### Dependencies
[Findings or "N/A - No new dependencies"]

### Best Practices
[Findings or "No issues found"]

### Code Duplication
[Findings or "No duplication detected"]

## Positive Observations ✅
[Note good practices and well-written code sections]

Severity Levels

  • 🔴 Critical: Must fix before merge (security issues, bugs, broken functionality)
  • 🟡 Warning: Should fix (performance, maintainability, potential issues)
  • 🔵 Suggestion: Consider fixing (style, minor improvements)

Instructions

  1. If reviewing a file, read the file first
  2. If reviewing a PR, use gh pr diff [number] to get the changes
  3. If reviewing staged changes, use git diff --cached
  4. If reviewing unstaged changes, use git diff
  5. For context, examine related files in the codebase
  6. Search for duplicated code patterns across the repository
  7. Check for existing linter/formatter configurations
  8. Provide specific line numbers for each finding
  9. Include code snippets showing the issue and suggested fix
  10. Be constructive and educational in feedback

Edge Cases

  • Binary files: Skip with note that binary files cannot be reviewed
  • Generated files: Note if file appears generated and skip detailed review
  • Empty diff: Report that there are no changes to review
  • Invalid file path: Ask user to provide a valid path
  • PR not found: Report error and ask for correct PR number
Skills similaires