Where are the review artifacts stored?

All reports, findings, and fix plans are stored in a `.code-review/` directory within your working folder, allowing you to audit the process and resume sessions if interrupted.

Can I choose which findings get fixed?

Yes. During the Adjudication stage, you interact with the Adjudication Lead to approve or reject specific findings before any code changes are planned or executed.

What makes this different from a standard /code-review?

While a standard review only lists findings, the Review Team skill orchestrates a multi-agent pipeline to adjudicate those findings with you, plan the necessary changes, and then execute the code fixes automatically.

What are the requirements to run this skill?

This skill requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS to be enabled, as it spawns four specialized agents (Review, Adjudication, Planning, and Execution Leads) to handle the workflow.

Is there a limit to how much code I can review at once?

For optimal quality, it is recommended to keep diffs under 2000 lines. Larger changes should be split into multiple review passes to ensure the agents remain accurate and thorough.

Review & Fix Agent Team

Name: Review & Fix Agent Team
Author: znorris

byznorris

0•

コラボレーションツール

Orchestrates a multi-stage pipeline of specialized agents to review code, adjudicate findings with the developer, and automatically implement approved fixes.

The Review Team skill transforms Claude into a Team Lead that manages a sophisticated four-agent pipeline designed for end-to-end code quality improvement. Instead of simply listing issues, it coordinates specialized Review, Adjudication, Planning, and Execution agents to systematically identify bugs, discuss findings with you, create a detailed remediation plan, and execute the actual code changes. This skill is particularly useful for complex pull requests or technical debt cleanup where you need a structured, autonomous process for both auditing and repairing code within a single workflow.

主な機能

01Four-stage specialized agent pipeline: Review, Adjudicate, Plan, and Execute

02Support for custom focus areas including security, performance, or API standards

03Human-in-the-loop adjudication to confirm, reject, or investigate findings

04Session resilience for resuming interrupted reviews and execution cycles

05Automated artifact generation in a local .code-review/ directory

060 GitHub stars

ユースケース

01Performing end-to-end audits and automated fixes for complex pull requests

02Validating specific architectural constraints against new feature implementations

03Cleaning up legacy files by applying modern best practices and error handling

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add znorris/claude-marketplace review-team

For use in Claude.ai and ChatGPT

主な機能

01Four-stage specialized agent pipeline: Review, Adjudicate, Plan, and Execute

02Support for custom focus areas including security, performance, or API standards

03Human-in-the-loop adjudication to confirm, reject, or investigate findings

04Session resilience for resuming interrupted reviews and execution cycles

05Automated artifact generation in a local .code-review/ directory

060 GitHub stars

ユースケース

01Performing end-to-end audits and automated fixes for complex pull requests

02Validating specific architectural constraints against new feature implementations

03Cleaning up legacy files by applying modern best practices and error handling