PR Review Toolkit: Six Review Subagents for Claude Code
pr-review-toolkit adds six Anthropic review subagents to Claude Code for comments, tests, silent failures, type design, code quality and simplification.
By Claudekit Team Published 1 min read
- Type
- Plugin
- Pricing
- Open Source
- Repository
- github.com
- Website
- github.com
pr-review-toolkit is an Anthropic plugin that bundles six review subagents, each focused on one aspect of a change. You can run them one at a time for a targeted check, or ask for several aspects before you open a pull request.
Checked 24 Oct 2026 against the plugin’s README, agent files and marketplace.json.
Who it is for
Developers who want a structured self-review before a PR, especially on code with new types, error handling or tests.
Install
/plugin install pr-review-toolkit@claude-plugins-officialThen ask in plain language, for example “Review the error handling in the API client”, or run the bundled /review-pr command.
What it does
The six agents:
- comment-analyzer: checks comments against the code and flags comment rot.
- pr-test-analyzer: rates test gaps from 1 to 10, focusing on behavior rather than line coverage.
- silent-failure-hunter: finds swallowed errors, weak catch blocks and bad fallbacks.
- type-design-analyzer: rates encapsulation and invariants on a 1 to 10 scale.
- code-reviewer: checks CLAUDE.md compliance and bugs, scoring issues 0 to 100.
- code-simplifier: reduces nesting and redundancy while keeping behavior the same.
Use cases
- Before committing: run code-reviewer and silent-failure-hunter.
- Before a PR: add pr-test-analyzer and type-design-analyzer.
- After review passes: let code-simplifier polish the result.
Pros
- Each agent has a narrow job, so findings are easier to act on.
- Works locally on any repo, no GitHub integration needed.
- Agents can run in parallel when you ask for several checks.
Cons and caveats
- Automatic agent selection depends on how you phrase the request.
- Running all six on a big diff costs a lot of tokens.
- The README says MIT while the plugin folder ships an Apache 2.0 LICENSE file; check the file if licensing matters to you.
Links
- Plugin folder: plugins/pr-review-toolkit
- Related: Code Review plugin, subagents guide, Claude Plugins Official
Frequently asked questions
How do I trigger an agent?
Ask for the kind of review you want, such as 'check for silent failures'. Claude picks the matching agent. You can also name the agent explicitly or use the review-pr command.
How is it different from the code-review plugin?
code-review posts one filtered comment on a GitHub PR. pr-review-toolkit gives you six specialist subagents you call locally, one aspect at a time or together.
Does it need GitHub?
No. The agents review code in your working tree, such as recent changes or a git diff. The review-pr command uses Bash, Glob, Grep, Read and Task tools.