What happens if the models disagree on a verdict?

The skill calculates a confidence score based on model agreement. If the score falls below your specified threshold, it returns an 'UNCLEAR' verdict, signaling that human review is required.

How does Council-Verify integrate with Git?

It uses snapshot IDs, such as Git commit SHAs, to pin the exact state of the code being verified. This ensures the audit trail is reproducible and accurately reflects the code at that point in time.

Can I use this for security audits?

Yes, the skill includes a rubric_focus parameter specifically for Security, allowing the council of models to prioritize vulnerability detection and compliance.

What is LLM Council multi-model deliberation?

It is a process where multiple AI models analyze the same code or document and deliberate to reach a consensus, ensuring more reliable and objective outcomes than a single model's assessment.

Council Verification & Multi-Model Validation

Name: Council Verification & Multi-Model Validation
Author: amiable-dev

byamiable-dev

•

보안 및 테스팅

Verifies code implementation and document accuracy using multi-model consensus to ensure high-quality, machine-actionable results.

The Council-Verify skill leverages the LLM Council framework to provide rigorous, multi-model deliberation for code and documentation. By synthesizing viewpoints from multiple AI models, it generates structured JSON verdicts with confidence scores, allowing developers to validate changes against specific rubrics like security, performance, or accessibility. This approach minimizes single-model bias and provides a reliable audit trail for complex implementation requirements, making it an essential tool for high-stakes AI governance and quality assurance.

주요 기능

01Structured JSON output with confidence thresholds and pass/fail verdicts

022 GitHub stars

03Customizable focus rubrics for Security, Performance, and Accessibility

04Multi-model consensus for objective and reliable code verification

05Git-integrated snapshot pinning for reproducible audit trails

06Automated CI/CD compatibility with standardized exit codes

사용 사례

01Enforcing quality gates in CI/CD pipelines using LLM-based deliberation

02Auditing complex technical documentation for completeness and accuracy

03Validating critical security patches before merging into production repositories

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add amiable-dev/conductor council-verify

For use in Claude.ai and ChatGPT

Download Skill