Can I target specific files or entire directories?

You can provide a specific file path, a directory, or even a focus area (e.g., 'security vulnerabilities' or 'data flow') to guide all four models toward a specific concern.

Does this skill work if I only have access to one or two of the models?

Yes, the skill features graceful degradation. It will identify which models are unavailable due to missing API keys or rate limits, notify you, and proceed to synthesize a report based on whichever models successfully completed their analysis.

What is the benefit of using four models for a single code review?

Using multiple models reduces 'hallucination' risks and individual model bias. It provides a consensus view where findings agreed upon by multiple models are highly likely to be valid, while unique insights from specific models (like Kimi's edge-case focus) provide deeper coverage.

Where are the review results stored?

Results are saved in a structured workspace directory (e.g., workspace/{task}QAQC/). Each model writes its own markdown report, and the orchestrator generates a final synthesis report in the same location.

Multi-Model QA/QC Reviewer

Name: Multi-Model QA/QC Reviewer
Author: gpt-cmdr

bygpt-cmdr

•

보안 및 테스팅

Orchestrates four independent AI models to perform parallel code reviews and synthesize a consensus report for critical bug investigation.

The Multi-Model QA/QC Reviewer is a specialized skill designed for high-stakes code validation and complex debugging. It launches four independent subagents—Opus, Gemini, Codex, and Kimi K2.5—to analyze targets from unique perspectives including architectural logic, large-context patterns, API contracts, and edge-case detection. By synthesizing individual findings into a single agreement matrix, the skill provides a 'wisdom of the crowd' approach to code quality, ensuring that critical vulnerabilities and subtle logic errors are caught through cross-validation while maintaining an audit trail of individual model reports.

주요 기능

0143 GitHub stars

02Parallel execution of four specialized AI models for independent analysis

03Automated synthesis of findings into a FINAL_QAQC_REPORT with an agreement matrix

04File-based workspace organization for easy auditing of individual model outputs

05Specialized focus areas including edge-case detection and code archaeology

06Graceful degradation logic that operates with 1 to 4 available models

사용 사례

01Investigating critical production bugs that require deep, multi-perspective reasoning

02Cross-validating complex logic in Jupyter notebooks or Python modules before major releases

03Performing thorough security and edge-case audits on sensitive API implementations

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add gpt-cmdr/ras-commander qa_review_triple-model

For use in Claude.ai and ChatGPT

Download Skill