Does this skill work with any programming language?

Yes, because it relies on git diffs and general programming principles, it can audit any language supported by Claude, including Go, TypeScript, Python, and Rust.

What happens if an agent finds an issue?

The main Claude agent will present the issues, fix them based on recommendations, and then re-trigger the review loop until all agents confirm the code is clean.

How many agents participate in the review?

The workflow uses three distinct sub-agents: one for Logic & Boundaries, one for Type Safety & Errors, and one for Concurrency & Resource Safety.

What is the benefit of the 'Writer-Reviewer' separation?

It prevents 'confirmation bias' where the AI that wrote the code overlooks its own logical flaws, ensuring a more critical and objective audit of the changes.

Claude Code Cross-Review

Name: Claude Code Cross-Review
Author: atompilot

byatompilot

0•

Seguridad y Pruebas

Implements a multi-agent review workflow that subjects code changes to three parallel expert audits before finalizing commits.

This skill formalizes the principle of 'the writer does not review' by deploying three specialized sub-agents to scrutinize code from different perspectives: logic and boundary conditions, type safety and error handling, and concurrency and resource management. By isolating these concerns and running reviews in parallel, it ensures high coverage and objectivity, forcing an iterative loop of fixes and re-reviews until the codebase is verified clean. It is designed to be triggered after development but before git commits, providing a rigorous automated quality gate for any software project.

Características Principales

01Prioritized Issue Reporting: Aggregates and ranks findings by severity, from logical flaws to resource management.

02Automated Git Integration: Streamlines the final commit process with standardized reports and co-authorship metadata.

03Parallel Expert Analysis: Simultaneously checks logic, type safety, and concurrency via specialized agents.

04Writer-Reviewer Separation: Eliminates self-review bias by delegating audits to independent sub-agents.

050 GitHub stars

06Iterative Quality Loop: Automatically re-evaluates fixes until all agents report a 'NO_ISSUES' status.

Casos de Uso

01Pre-commit validation for complex feature development to ensure no regressions or edge-case failures.

02Ensuring high-stakes code, such as concurrency logic or financial calculations, is audited by multiple specialized AI perspectives.

03Automating rigorous code review standards in solo development environments where a second pair of eyes is unavailable.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add atompilot/claude-code-cross-review claude-review

For use in Claude.ai and ChatGPT

Características Principales

01Prioritized Issue Reporting: Aggregates and ranks findings by severity, from logical flaws to resource management.

02Automated Git Integration: Streamlines the final commit process with standardized reports and co-authorship metadata.

03Parallel Expert Analysis: Simultaneously checks logic, type safety, and concurrency via specialized agents.

04Writer-Reviewer Separation: Eliminates self-review bias by delegating audits to independent sub-agents.

050 GitHub stars

06Iterative Quality Loop: Automatically re-evaluates fixes until all agents report a 'NO_ISSUES' status.

Casos de Uso

01Pre-commit validation for complex feature development to ensure no regressions or edge-case failures.

02Ensuring high-stakes code, such as concurrency logic or financial calculations, is audited by multiple specialized AI perspectives.

03Automating rigorous code review standards in solo development environments where a second pair of eyes is unavailable.