Comprehensive PR Review Workflow

Run multi-agent pull request reviews where each specialized agent focuses on a different aspect of code quality, then aggregate results into an actionable summary.

Review Aspects

Six specialized review agents are available:

Review Workflow

Step 1: Determine Review Scope

1. Check git status to identify changed files: git diff --name-only
2. Check if a PR already exists: gh pr view
3. Parse any user-specified review aspects
4. Default: run all applicable reviews

Step 2: Identify Applicable Reviews

Based on the changes:

• Always applicable: code-reviewer (general quality)
• If test files changed: pr-test-analyzer
• If comments/docs added: comment-analyzer
• If error handling changed: silent-failure-hunter
• If types added/modified: type-design-analyzer
• After passing review: code-simplifier (polish and refine)

Step 3: Launch Review Agents

Sequential approach (default):

• One agent at a time for easier understanding and action
• Each report is complete before the next
• Good for interactive review

Parallel approach (on user request):

• Launch all agents simultaneously
• Faster for comprehensive review
• Results come back together

Step 4: Aggregate Results

After agents complete, organize findings:

# PR Review Summary

## Critical Issues (X found)
- [agent-name]: Issue description [file:line]

## Important Issues (X found)
- [agent-name]: Issue description [file:line]

## Suggestions (X found)
- [agent-name]: Suggestion [file:line]

## Strengths
- What's well-done in this PR

## Recommended Action
1. Fix critical issues first
2. Address important issues
3. Consider suggestions
4. Re-run review after fixes

Agent Details

code-reviewer

Reviews code against project guidelines in CLAUDE.md with confidence-based filtering (only reports issues with confidence >= 80). Checks for CLAUDE.md compliance, bug detection, and code quality. Groups issues by severity (Critical: 90-100, Important: 80-89).

pr-test-analyzer

Focuses on behavioral coverage rather than line coverage. Identifies critical code paths, edge cases, and error conditions that must be tested. Rates criticality from 1-10 and maps to severity (9-10: CRITICAL, 7-8: HIGH, 5-6: MEDIUM, 1-4: LOW).

silent-failure-hunter

Analyzes for silent failures, empty catch blocks, broad exception catching, inadequate error messages, and unjustified fallback behavior. Checks logging quality, user feedback, catch block specificity, and error propagation.

type-design-analyzer

Evaluates new types on four dimensions (each rated 1-10): encapsulation, invariant expression, invariant usefulness, and invariant enforcement. Flags anti-patterns like anemic domain models, exposed mutable internals, and missing construction validation.

comment-analyzer

Verifies comment factual accuracy against code, assesses completeness, evaluates long-term value, and identifies misleading elements. Categorizes findings as critical issues, improvement opportunities, and recommended removals.

code-simplifier

Simplifies code for clarity and maintainability while preserving all functionality. Applies project standards, reduces unnecessary complexity, eliminates redundant code, and improves naming. Avoids nested ternaries and over-compact solutions.

Usage Patterns

Full review (default): Run all applicable review agents based on changed files.

Targeted review: Specify aspects to focus on: tests, errors, comments, types, code, simplify.

Parallel review: Request all agents run simultaneously for faster results.

Workflow Integration

Before committing:

1. Write code
2. Run code and error reviews
3. Fix critical issues
4. Commit

Before creating PR:

1. Stage all changes
2. Run all reviews
3. Address critical and important issues
4. Re-run specific reviews to verify
5. Create PR

After PR feedback:

1. Make requested changes
2. Run targeted reviews based on feedback
3. Verify issues are resolved
4. Push updates

Tips

• Run reviews early, before creating the PR
• Agents analyze git diff by default
• Address critical issues before lower priority ones
• Re-run reviews after fixes to verify resolution
• Use targeted reviews when the concern area is known
• All agents return actionable results with specific file:line references