agentskills.codes
CO

code-review

Review code for bugs, security issues, and quality. Use when asked to review code or before creating a PR.

Install

mkdir -p .claude/skills/code-review-james10192 && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16490" && unzip -o skill.zip -d .claude/skills/code-review-james10192 && rm skill.zip

Installs to .claude/skills/code-review-james10192

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Review code for bugs, security issues, and quality. Use when asked to review code or before creating a PR.
106 chars✓ has a “when” trigger

About this skill

Perform a thorough code review of the changes in the current branch.

Step 1 — Get the changes

git diff develop...HEAD
git diff --stat develop...HEAD

If there is no develop branch, use main or master as the base.

Step 2 — Review checklist

Security (CRITICAL — block PR if any found)

  • No hardcoded secrets, tokens, or credentials in code
  • No SQL string interpolation (use parameterized queries / ORM)
  • User input validated before processing
  • Authentication/authorization enforced on protected routes
  • No sensitive data exposed in API responses (passwords, tokens)
  • No console.log / print of sensitive data

Architecture & Patterns

  • Code follows the project's existing conventions (check CLAUDE.md)
  • Business logic separated from HTTP/presentation layer
  • No circular dependencies introduced
  • Functions do one thing (single responsibility)

Code Quality

  • Functions < 50 lines
  • No duplicated logic (DRY)
  • Meaningful variable and function names
  • No commented-out code left behind
  • Proper error handling (no silent failures, no bare catch / except)
  • No unnecessary any types (TypeScript projects)

Tests

  • New features have tests
  • Bug fixes have regression tests
  • Happy path + main error cases covered

Performance

  • No obvious N+1 queries
  • No blocking operations in async contexts
  • Large lists paginated, not fetched all at once

Step 3 — Report

Format your review as:

BLOCKING — Must fix before merge:

  • [issue] in file:line

IMPORTANT — Should fix:

  • [issue] in file:line

SUGGESTIONS — Nice to have:

  • [suggestion]

GOOD — What was done well

$ARGUMENTS

Search skills

Search the agent skills registry