How does this skill handle overlapping review concerns?

The skill defines strict scope boundaries. For example, it distinguishes between 'Types' (logic/states) and 'Consistency' (naming/patterns) to ensure unique and relevant feedback.

How are different AI models allocated for reviews?

Models are assigned based on 'Reasoning Demand'. High-reasoning tasks like architectural implications use Claude 3 Opus, while pattern-matching tasks like documentation checks use Sonnet.

What are the priority tiers in this catalog?

Agents are ranked by criticality. Tier 1 includes security (IAM) and architecture because misses in these areas cause irreversible debt or incidents, whereas Tier 3 covers recoverable issues like missing tests.

What is the PR Review Subagent Catalog?

It is a structured directory of specialized AI agents designed to review specific aspects of code, such as security, architecture, and frontend patterns, ensuring each area gets expert-level attention.

Is this skill meant to be invoked directly by users?

No, this is a system-level skill (user-invocable: false) designed to guide orchestrator agents in dispatching the correct subagents during a PR review process.

PR Review Subagent Catalog

Name: PR Review Subagent Catalog
Author: inkeep

byinkeep

•

コラボレーションツール

Orchestrates specialized AI subagents to perform deep, domain-specific code reviews across architecture, security, and performance.

This skill acts as a central registry and routing guide for a suite of specialized PR review subagents. It defines the specific domains—ranging from IAM security and architectural integrity to product impact and error handling—that different AI models manage during the automated review process. By categorizing review tasks based on reasoning demand and criticality, it ensures that high-stakes issues are handled by top-tier models like Claude 3 Opus, while standard pattern-matching tasks are efficiently processed by Sonnet, providing a comprehensive and cost-effective review ecosystem.

主な機能

01Defined scope boundaries to eliminate overlapping or redundant comments

02Tiered priority ranking for high-stakes security and architecture checks

03Model allocation mapping based on reasoning demand and criticality

043 GitHub stars

05Catalog of 17+ specialized subagents for granular code analysis

06Quick-reference pattern matching for common code review scenarios

ユースケース

01Routing code review tasks to appropriate AI models based on complexity

02Orchestrating automated multi-agent PR review workflows in CI/CD

03Defining clear ownership areas for AI-driven code quality standards

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add inkeep/team-skills pr-review-subagents-available

For use in Claude.ai and ChatGPT

Download Skill