Agent-Skills.md

Agent Skills: Review - Comprehensive Code Review

Conduct comprehensive code reviews using 5 specialized subagents (code quality, performance, test coverage, documentation accuracy, security) in parallel, providing an integrated report. Use at important milestones (5+ file changes, public API updates, infrastructure config updates, before commits/PRs/releases). Triggers on requests like "review this", "check the code", "review the changes".

UncategorizedID: ohbarye/dotfiles/review

Repository

ohbarye/dotfiles
ohbarye
6

Install this agent skill to your local

pnpm dlx add-skill https://github.com/ohbarye/dotfiles/tree/HEAD/agents/skills/review

Skill Files

Browse the full folder contents for review.

Download Skill

Loading file tree…

agents/skills/review/SKILL.md

Skill Metadata

Name
review
Description
Conduct comprehensive code reviews using 5 specialized subagents (code quality, performance, test coverage, documentation accuracy, security) in parallel, providing an integrated report. Use at important milestones (5+ file changes, public API updates, infrastructure config updates, before commits/PRs/releases). Triggers on requests like "review this", "check the code", "review the changes".

Review - Comprehensive Code Review

Conduct multi-perspective code reviews using specialized subagents.

When to Use This Skill

Automatically suggest a review in the following situations:

  • User says "review this", "check the code", or "review the changes"
  • After completing an implementation involving 5+ file changes
  • After public API updates
  • Before creating a commit or PR
  • User runs /review

Review Perspectives

| Subagent | Perspective | Focus | |---|---|---| | code-quality-reviewer | Code Quality | Clean code, SOLID principles, readability, maintainability | | performance-reviewer | Performance | Bottlenecks, N+1 queries, memory efficiency | | test-coverage-reviewer | Test Coverage | Completeness, test quality, boundary values | | documentation-accuracy-reviewer | Documentation Accuracy | Code-documentation alignment | | security-code-reviewer | Security | OWASP Top 10, input validation, authentication & authorization |

Execution Flow

Step 1: Identify Review Targets

  1. Use git diff and git diff --staged to check changes
  2. If changes are extensive, list changed files to understand the review scope

Step 2: Parallel Subagent Execution

IMPORTANT: Execute all 5 subagents in parallel.

Provide each subagent with:

  • Diffs of changed files
  • Relevant context (existing code, design intent)
  • Instructions to report only noteworthy findings

Step 3: Create Integrated Report

Once all reviews are complete, synthesize feedback into the following format:

  1. Summary: Concise assessment of overall change quality
  2. Critical Issues (if any): Issues requiring immediate attention
  3. Improvement Suggestions (if any): Suggestions for quality enhancement
  4. Positive Highlights (if any): Excellent implementations and good practices

Step 4: Review Cycle

If issues are found:

  1. Fix the issues
  2. Run /review again
  3. Repeat until no issues remain

Conflict Resolution

When reviewers disagree or a design trade-off is unclear, use /magi-decision-support to get multi-perspective analysis. The MAGI system provides three viewpoints (Scientist, Mother, Realist) that can help resolve:

  • Contradicting recommendations from different reviewers
  • Architectural trade-offs with no clear winner
  • Performance vs. readability vs. security trade-offs
  • Any decision where reasonable engineers would disagree

Notes

  • Keep feedback concise and focus on important issues only
  • Avoid burying critical problems under trivial concerns
  • When code is well-written, acknowledge it explicitly
Review - Comprehensive Code Review Skill | Agent Skills