requesting-code-review

requesting-code-review - Code Review Skill Details

Skill Overview

requesting-code-review is a code review skill for Claude Code. By dispatching the superpowers:code-reviewer subagent, it helps developers identify and fix issues early after completing development tasks, implementing major features, or verifying that work meets requirements before merging.

Applicable Scenarios

1. Subagent-Driven Development Workflows

Conduct a code review immediately after each subagent task is completed to prevent issues from accumulating in subsequent tasks. This principle of “review early, review often” helps ensure the quality of each stage and avoid the cost of rework.

2. Validation After Completing Major Features

After completing a core function or significant feature, use this skill for a comprehensive review to ensure that the implementation meets the planned requirements, the functional logic is correct, and the code quality meets the expected standards.

3. Quality Assurance Before Merging

Conduct a mandatory code review before merging code into the main branch to verify the correctness, security, and maintainability of the changes, thereby reducing the risk of production issues.

Core Features

1. Intelligent Subagent Dispatch

Use the Task tool to dispatch a subagent of type superpowers:code-reviewer, automatically providing the reviewer with precise contextual information. The reviewer receives carefully designed evaluation context rather than your session history, ensuring that the review focuses on the work itself while preserving your own context so you can continue working.

2. Automatic Git SHA Extraction

Provides standard Bash commands for quickly obtaining the base SHA and current SHA, supporting precise code reviews based on commit ranges. You can specify a particular base commit, such as HEAD~1 or origin/main, to ensure that the review covers the correct range of code changes.

3. Templated Review Workflow

Provides a complete code-reviewer.md template containing key placeholders—including implementation details, planned requirements, base SHA, current SHA, and a description—to simplify the preparation of review requests and ensure that every review includes the necessary information.

Frequently Asked Questions

When should a code review be conducted?

Mandatory scenarios include after every task in subagent-driven development, after completing a major feature, and before merging into the main branch. Recommended scenarios include when facing development challenges (to gain a new perspective), before refactoring (to establish a baseline), and after fixing complex bugs. The core principle is “review early, review often.”

How do I use the requesting-code-review skill?

First, obtain the Git SHAs (BASE_SHA and HEAD_SHA), then use the Task tool to dispatch the superpowers:code-reviewer subagent. Fill in the placeholders in the template, including the implementation details, planned requirements, commit range, and a brief description. After the review is complete, take the appropriate action based on the feedback level: fix Critical issues immediately, fix Important issues before continuing, and record Minor issues for later handling.

What should I do if the code review identifies issues?

Take different actions based on the severity of each issue: Critical issues must be fixed immediately; Important issues must be fixed before proceeding to the next step; Minor issues can be documented and addressed at an appropriate time. If you believe the reviewer’s judgment is incorrect, provide technical reasoning, demonstrate the code or test evidence supporting its validity, or request clarification.