How does the verification step improve report quality?

After subagents identify potential issues, the system cross-references every finding with the actual code to ensure the issue is real and not an AI misinterpretation.

Does sd-review modify my code automatically?

No, sd-review is an analysis-only skill. it provides comprehensive reports and suggestions but does not make any changes to your source files.

Can I target specific directories for review?

Yes, you can specify a path like '/sd-review packages/core' to focus the analysis on a specific module or package.

When does the security-specific reviewer run?

The security agent is automatically triggered if the initial architectural analysis detects ORM queries, database builders, or service endpoints.

What is the difference between this and a standard linter?

While linters check for syntax and style, sd-review uses multiple AI agents to find deep logical errors, security vulnerabilities, and architectural inconsistencies.

Multi-Perspective Code Reviewer

Name: Multi-Perspective Code Reviewer
Author: kslhunter

bykslhunter

0•

セキュリティとテスト

Performs deep, multi-agent code analysis to identify bugs, security risks, and architectural improvements without modifying source code.

The sd-review skill provides an exhaustive, multi-perspective analysis of your codebase using specialized parallel subagents. It begins by mapping out your project's architecture through an Explore agent, then dispatches reviewers focused on logic, complexity, API design, and security. Unlike simple linters, it verifies every finding against the source code to eliminate false positives, delivering a structured, prioritized report. It is an essential tool for comprehensive pre-merge checks, high-stakes quality audits, and facilitating faster developer onboarding by providing a high-level overview of complex directories.

主な機能

01Structured reporting with prioritized P0-P4 severity levels

02Parallel subagent analysis covering logic, security, and DX

03Automated architectural mapping via the built-in Explore agent

04Conditional security auditing for ORM queries and service endpoints

05Mandatory verification step to eliminate AI false-positives

060 GitHub stars

ユースケース

01Onboarding to unfamiliar codebases by generating a quality and architecture overview

02Performing a final safety audit before merging major features or refactors

03Identifying technical debt and simplification opportunities in legacy packages

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kslhunter/simplysm sd-review

For use in Claude.ai and ChatGPT

Download Skill