code-review

Detect and document code quality issues in Python/PyTorch code.

2|1|Updated Jan 8, 2026

Search Tags:#high-performance #pytorch #code-review #software-engineering #production-readiness

Ready to run

npx skills add https://github.com/biobenkj/flash-semicrf --skill code-review-biobenkj

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/biobenkj/flash-semicrf/tree/main/.claude/skills/code-review
Command: npx skills add https://github.com/biobenkj/flash-semicrf --skill code-review-biobenkj

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Systematic, reproducible code reviews for high-performance Python/PyTorch projects, helping teams improve correctness, scalability, and production readiness.

Core Features & Use Cases

Comprehensive review checklists covering correctness, memory usage, numerical stability, and PyTorch/CUDA practices.
Actionable, severity-based reporting with concrete code improvement suggestions and examples.
Useful for PR audits, onboarding reviews, and pre-release production readiness checks.

Quick Start

Review the files in the PR and report findings by severity with actionable recommendations.

Frequently Asked Questions about code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I review PyTorch code for production readiness and memory usage?▼

Review PyTorch code for production readiness by analyzing memory usage, device placement, and numerical stability. This process identifies edge cases and scalability issues, providing actionable recommendations with code examples aligned to PyTorch best practices.

What is the best way to detect numerical stability and correctness issues in PyTorch projects?▼

Detect numerical stability and correctness issues in PyTorch projects by applying a comprehensive code review checklist. This analyzes high-performance Python code to document edge cases and ensure scalability across streaming and exact backends.

How do I check CUDA device placement and memory efficiency in my Python code?▼

Check CUDA device placement and memory efficiency in Python code by performing a systematic review. This evaluates PyTorch and CUDA practices, reporting severity-based findings with concrete suggestions to improve overall code quality.

Does this code review process support Semi-Markov CRF conventions and streaming backends?▼

Yes, this code review process supports Semi-Markov CRF conventions and streaming backends. It analyzes code across both streaming and exact backends, ensuring correctness and scalability considerations are thoroughly documented.

Can I use this for pre-release PR audits on high-performance Python projects?▼

Yes, you can use this for pre-release PR audits on high-performance Python projects. It provides systematic, reproducible code reviews that generate severity-based reports with actionable code improvement suggestions.

Why does my PyTorch code fail scalability checks during onboarding reviews?▼

PyTorch code fails scalability checks during onboarding reviews due to unoptimized memory usage, incorrect device placement, or unhandled edge cases. A rigorous review identifies these issues and provides aligned recommendations for improvement.