Claude Commands - Review

review.md

description
AI code review using specialized subagents (/review)

Perform comprehensive code review using specialized AI agents working in parallel

IMPORTANT: Never modify files - only output review files

Execution Conditions

You must verify the following conditions before proceeding:

• Pull Request exists for the current branch (draft or opened)
• Template file exists at ~/obsidian/templates/review.md

If any condition is not met:

• Stop the process immediately
• Notify the user which condition failed
• Do not proceed with implementation

Specialized Review Agents

The review system employs 8 domain-specific agents:

1. Product Manager: Requirements alignment, acceptance criteria
2. Backend Developer: Server-side architecture, APIs
3. Frontend Developer: UI components, state management, performance, cache strategies
4. UI Engineer: Web standards consistency for UI
5. Infrastructure Engineer: AWS, Terraform
6. Database Engineer: MySQL, PostgreSQL, ORMs
7. Quality Engineer: Testing, reliability, maintainability
8. Refactoring Expert: Code quality, naming conventions, design principles

Execution Process

When all conditions are met, execute these phases in order:

Phase 1: Verify Pull Request Existence and Get Information

# Get current branch (already checked out to Pull Request branch)
current_branch=$(git branch --show-current)

# Check if Pull Request exists for current branch
pr_number=$(gh pr list --head "$current_branch" --state all --json number --jq '.[0].number')

# Verify Pull Request exists
if [ "$pr_number" = "null" ] || [ -z "$pr_number" ]; then
    exit 1
fi

# Get base branch from Pull Request information
base_branch=$(gh pr view "$pr_number" --json baseRefName --jq '.baseRefName')

Phase 2: Collect Diff

# Get changed files and diff
changed_files=$(git diff --name-only ${base_branch}...HEAD)
git_diff=$(git diff ${base_branch}...HEAD)

Phase 3: Select Agents

Analyze the context and changes to determine which specialized subagents to invoke:

1. Analyze Changed Files and Directories

   • Examine file extensions, paths, and directory structure
   • Review git diff content to understand the nature of changes

2. Determine Required Agents

   Based on the analysis, select relevant subagents:

   • Always invoke: pm-reviewer, qa-reviewer, refactoring-reviewer
   • Conditionally invoke based on changes:
     • backend-reviewer: When changes in any backend files
     • frontend-reviewer: When changes any in frontend files
     • ui-reviewer: When changes in any frontend files
     • infrastructure-reviewer: When changes in any infrastructure files

Phase 4: Parallel Agent Execution

1. Launch selected subagents in parallel

Let subagents review for domain-specific areas.

2. Check execution

Verify that all the selected subagents are running. If there is any subagent not running, be sure to execute them.

Phase 5: Result Aggregation

1. Collect Agent Outputs

   • Wait for all the selected subagents to complete
   • Gather findings from each output file (e.g. ./tmp/pm-review.md or ./backend-review.md)

2. Categorize Findings

   Group all findings by severity:

   • ❌ Critical Issues: Must-fix problems
   • ⚠️ Minor Issues: Non-critical improvements
   • ℹ️ Info & Questions: Clarifications needed

3. Maintain Attribution

   • Preserve which agent provided each finding
   • Format: [Agent Name] Finding description (file:line)

Phase 6: Generate Unified Report

Create comprehensive review report in ./tmp/review.md. Structure your findings in the following format: ~/obsidian/templates/review.md.

Phase 7: Completion Report

Provide summary to user:

• Critical issues to do the next
• Output files