Agent Plugin
PR Review Toolkit
Comprehensive collection of six specialized review agents that analyze pull requests for comment accuracy, test coverage, error handling, type design, code quality, and simplification opportunities.
What is PR Review Toolkit?
The PR Review Toolkit provides deep, specialized code review capabilities through six expert agents that each focus on a distinct quality dimension of pull requests. Rather than offering generic code review feedback, each agent performs targeted analysis with specific evaluation criteria and actionable recommendations. The comment-analyzer examines documentation accuracy against actual code behavior and identifies comment rot and technical debt in inline documentation. The pr-test-analyzer evaluates test coverage quality by analyzing behavioral coverage rather than just line coverage, identifying critical testing gaps and edge cases that lack validation. The silent-failure-hunter specifically targets error handling quality, catching inadequate error handling, inappropriate fallback behavior, and missing error logging that could hide failures in production. The type-design-analyzer rates type designs across four dimensions including encapsulation strength, invariant expression, type usefulness, and invariant enforcement with numerical scores from one to ten. The code-reviewer performs general quality assessment against project guidelines defined in CLAUDE.md files, checking style compliance and detecting potential bugs. The code-simplifier identifies opportunities to reduce complexity without changing behavior. Teams can invoke individual agents for targeted reviews or run multiple agents together for comprehensive pull request analysis before merging.
Key capabilities
Specialized Review Agent Architecture
Provides six independent expert agents that each focus on a single quality dimension, allowing developers to request targeted reviews for specific concerns or run comprehensive multi-agent analysis across all quality dimensions simultaneously.
Comment Accuracy and Documentation Analysis
The comment-analyzer agent evaluates whether inline comments and documentation accurately reflect actual code behavior, identifying outdated comments that describe previous implementations, misleading documentation, and comment rot that creates technical debt.
Behavioral Test Coverage Assessment
The pr-test-analyzer examines test quality beyond simple line coverage metrics, analyzing whether tests validate actual behavioral requirements, cover critical edge cases and error conditions, and demonstrate resilience to implementation changes.
Quantified Type Design Evaluation
The type-design-analyzer provides numerical ratings from one to ten across four specific dimensions including type encapsulation strength, invariant expression clarity, type usefulness, and invariant enforcement mechanisms, enabling objective comparison of type design quality.
Use cases
Pre-Merge Pull Request Quality Gates
Before merging significant pull requests, development teams run the full toolkit to get comprehensive feedback across all six quality dimensions, catching issues that might be missed in manual review and ensuring consistent quality standards across the codebase.
Targeted Review for Specific Code Quality Concerns
When a pull request modifies error handling logic, developers invoke the silent-failure-hunter agent specifically to ensure all catch blocks properly log errors and handle failures appropriately, or use the comment-analyzer when documentation changes need verification.
Onboarding New Contributors to Project Standards
New team members submit pull requests that are automatically reviewed by the code-reviewer agent against the project's CLAUDE.md guidelines, receiving immediate feedback on style violations and coding standards without requiring senior developer time.
Type System Refactoring Validation
During refactoring efforts that introduce new type abstractions, developers use the type-design-analyzer to evaluate whether the new types properly encapsulate invariants and express constraints, comparing numerical scores across different design alternatives.
Included components
Commands
PR Review Toolkit
Agents
Specialist reviewers