Can I tell the AI to focus on specific issues like security?

Yes, you can use the 'customInstructions' argument to provide specific guidance, such as 'focus on security' or 'ensure this follows our internal naming conventions.'

Does this skill require an external server to work?

No, the analyze skill works in standalone mode. While it can push results to the Pair Review web UI if it's running, it performs all analysis using local scripts and parallel task agents.

What are the three levels of analysis performed?

Level 1 focuses on the diff isolation, Level 2 examines the full file context, and Level 3 analyzes the broader codebase architecture and dependencies.

How do the 'tier' options affect the review?

The tiers (fast, balanced, thorough) determine the complexity of the AI models used. 'Fast' uses lighter models for quick checks, while 'Thorough' utilizes deep-reasoning models for complex architectural reviews.

AI Code Review Critic

Name: AI Code Review Critic
Author: in-the-loop-labs

byin-the-loop-labs

•

协作工具

Performs multi-layered AI code reviews using parallel agents to identify bugs, improvements, and architectural issues across your codebase.

The AI Code Review Critic (plugin-code-critic) provides a sophisticated, three-tiered analysis of your code changes by spawning parallel task agents directly within your Claude Code environment. It evaluates changes at the isolation level, within full file contexts, and across the broader codebase architecture to deliver curated, high-confidence suggestions. Whether used as a standalone tool or integrated with the Pair Review web UI, this skill streamlines the review process by deduplicating insights and providing actionable fixes tailored to your specific repository patterns.

主要功能

01Seamless orchestration that merges and deduplicates suggestions into a final report

02Configurable analysis tiers including fast, balanced, and thorough processing

03Three-level contextual analysis covering diffs, file context, and codebase architecture

04Support for custom review instructions to enforce specific team patterns or security goals

05Parallel subagent execution for fast, multi-perspective code evaluation

0616 GitHub stars

使用场景

01Visualizing AI suggestions inline using the optional Pair Review web interface

02Enforcing domain-specific best practices and security standards through custom instructions

03Automating deep code reviews for local changes before opening a pull request

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add in-the-loop-labs/pair-review analyze

For use in Claude.ai and ChatGPT

Download Skill