llm-review

Run external LLM reviews across Gemini and OpenAI with consistent JSON outputs.

Updated Apr 5, 2026

Search Tags:#consistency #llm #gemini #openai #retry #wrapper #json-output

Ready to run

npx skills add https://github.com/TeamSPWK/nova-algorithm --skill llm-review-teamspwk

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: llm-review
Source: https://github.com/TeamSPWK/nova-algorithm/tree/main/skills/llm-review
Command: npx skills add https://github.com/TeamSPWK/nova-algorithm --skill llm-review-teamspwk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates and standardizes external LLM reviews across Gemini and OpenAI to ensure consistent outputs and traceable reasoning.

Core Features & Use Cases

Uniform API and JSON outputs across providers for easy integration.
Enforces fixed model configurations to prevent agent modification and promote reproducibility.
Use cases include code review, regulatory thinking, and content auditing with auditable reasoning.

Quick Start

Prompt the system to compare inputs using Gemini and GPT and return a unified JSON result.

Frequently Asked Questions about llm-review

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

FAQPage Schema

How do I standardize JSON output from multiple LLMs like Gemini and OpenAI?▼

To standardize JSON output across Gemini and OpenAI, a wrapper enforces fixed model mappings and consistent output fields. This ensures uniform API responses and traceable reasoning across different providers for evaluation tasks.

Can I use fixed model configurations for code review to prevent agent modification?▼

Yes, fixed model configurations can be enforced for code review to prevent agent modification. This approach promotes reproducibility by standardizing the external LLM review process and ensuring the evaluation logic remains controlled.

What is the best way to handle API errors and retries when running multi-LLM reviews?▼

The best way to handle API errors during multi-LLM reviews is implementing standardized error handling and retry policies within your wrapper. This maintains reliability when querying Gemini and OpenAI for structured evaluation tasks.

Does multi-LLm review support content auditing with optional source enrichment?▼

Yes, multi-LLM review supports content auditing by reading optional sources to enrich prompts. This allows deeper regulatory thinking and auditable reasoning while maintaining a consistent JSON output format across providers.

Why does my LLM wrapper return inconsistent results across different providers?▼

Your LLM wrapper returns inconsistent results across providers due to varying model behaviors and lack of enforced output structures. Implementing fixed model mappings and standardized JSON output fields resolves this by enforcing uniform response schemas.

When do I need a unified wrapper for external LLM reviews?▼

You need a unified wrapper for external LLM reviews when performing structured evaluation tasks like code review, regulatory thinking, or content auditing. It ensures consistent JSON outputs, error handling, and reproducible reasoning across Gemini and OpenAI.