3-Agent Parallel Code Review Skill for Claude Code

The 3‑Agent Parallel Code Review Skill adds three focused review passes to Claude Code’s workflow, letting you catch reuse problems, quality issues, and efficiency bottlenecks in a single command. By spawning dedicated agents for each concern, it aggregates findings, removes duplicates, and produces a severity‑ranked report that can be applied automatically for the highest‑impact fixes.

What You Get

• /simplify Skill – Executes the full three‑agent parallel review. It determines the review scope from the current git diff, launches the Code Reuse, Code Quality, and Efficiency agents simultaneously, merges their outputs, filters false positives, and returns a concise report grouped by severity (HIGH, MEDIUM, LOW). An optional --fix flag auto‑applies fixes for HIGH severity items.
• /review Skill – Runs a targeted review on a specific file, directory, or pull‑request number. It fetches the PR diff via gh pr diff and applies the same three‑agent analysis only to the selected scope.
• Code Reuse Agent – Detects duplicated logic, suggests existing utilities, flags redundant patterns, and highlights copy‑pasted code that varies only slightly.
• Code Quality Agent – Checks naming consistency, compliance with CLAUDE.md guidelines, common code smells, magic numbers, deep nesting, and unnecessary complexity.
• Efficiency Agent – Looks for N+1 query patterns, superfluous re‑renders, missed concurrency opportunities, memory‑leak risks, and code that inflates bundle size.

How to Use

• Copy the skill files into your Claude configuration:
  cp -r skills/* .claude/skills/
cp commands/*.md .claude/commands/
• Build or modify your feature branch and run your test suite.
• Execute /simplify to run a full parallel review of all changes staged for commit.
• If you only need to inspect a particular PR, use /review <pr-number> (or a file path) to limit the analysis.
• Review the generated report. HIGH severity items can be fixed automatically with /simplify --fix, while MEDIUM and LOW items are presented for manual decision.
• Commit the cleaned code and push.

Who This Is For

• Solo developers who want a quick, comprehensive sanity check before committing.
• Engineers maintaining legacy codebases where hidden duplication and inefficiency are common.
• Teams that rely on Claude Code for automated assistance but need an extra layer of validation without writing custom lint rules.
• Anyone who prefers a single command that surfaces reuse, quality, and performance concerns together, rather than running separate tools.