Can I use only one model instead of both?

Yes, you can specify 'codex' or 'gemini' in your command arguments to limit the review to a single external provider.

Is my code sent to external services?

Yes, relevant code context or diffs are sent to OpenAI and Google APIs to perform the requested analysis. Secrets and credentials are automatically excluded from these prompts.

Do I need to have Codex and Gemini installed?

Yes, this skill requires the Codex MCP and the Gemini CLI tool to be configured in your environment to successfully delegate tasks.

What happens if one of the external models is down?

The skill includes robust error handling that will proceed with the available model and inform you of the specific failure for the missing tool.

How does this skill differ from Claude's standard review?

While Claude is highly capable, this skill provides independent validation from different model families (OpenAI and Google), reducing bias and increasing the likelihood of catching edge cases.

Second Opinions

Name: Second Opinions
Author: vinta

byvinta

•

112

•

Security & Testing

Gathers independent perspectives from OpenAI Codex and Google Gemini to provide multi-model reviews of code, architecture, and documentation.

The Second Opinions skill empowers Claude to act as a coordinator for cross-model validation, delegating complex tasks like code reviews, architectural planning, and documentation audits to external models like OpenAI Codex and Google Gemini. By automating context gathering—including Git diffs, project conventions from CLAUDE.md, and codebase analysis—it provides a synthesized report that highlights consensus, identifies critical issues, and offers diverse perspectives from different AI model families to ensure higher code quality and more robust system designs.

Key Features

01Iterative feedback loops for architectural and planning tasks

02Multi-model synthesis from OpenAI Codex and Google Gemini

03Parallel dispatching of external model requests for faster results

04Support for specialized focus areas like security, performance, and correctness

05Automated Git diff and context extraction for comprehensive reviews

06112 GitHub stars

Use Cases

01Performing a secondary code review before merging a critical pull request

02Analyzing a codebase for subtle race conditions or pattern inconsistencies

03Validating a complex system migration plan from multiple AI perspectives

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add vinta/hal-9000 second-opinions

For use in Claude.ai and ChatGPT

Download Skill