What is LLM Council Verification?

It is a Claude Code skill that uses multiple AI models to deliberate and reach a consensus on the quality, completeness, and correctness of code or documentation.

How does the confidence threshold work?

Users can set a minimum confidence level; if the multi-model consensus score falls below this threshold, the skill returns an 'UNCLEAR' verdict, signaling that human review is required.

What kind of rubrics can I apply?

You can focus the verification process on specific domains such as Security, Performance, or Accessibility to ensure the models prioritize those criteria during deliberation.

Can I use this in my CI/CD pipeline?

Yes, the skill provides specific exit codes (0 for PASS, 1 for FAIL, 2 for UNCLEAR) specifically designed for integration into automated deployment and quality gate workflows.

Does it keep a record of the verification process?

Yes, it creates a persistent audit trail and stores transcripts of the deliberation, providing full transparency into how the AI reached its verdict.

LLM Council Verification

Name: LLM Council Verification
Author: amiable-dev

byamiable-dev

•

セキュリティとテスト

Verifies code, documentation, and implementations through multi-model consensus to ensure accuracy and compliance with requirements.

概要

LLM Council Verification is a high-assurance tool that leverages the deliberation of multiple AI models to provide rigorous, structured verdicts on code quality, security, and implementation accuracy. By aggregating perspectives from various LLMs, it produces machine-actionable results—complete with confidence scores and audit trails—that go beyond simple linting. This skill is essential for developers and teams who require high-confidence validation for critical changes, ensuring that every commit or document meets specific rubrics before being finalized.

主な機能

Structured JSON output with scores and blocking issues
Snapshot-based reproducibility using Git commit SHAs
Multi-model consensus for high-confidence pass/fail verdicts
Automated audit trails for transparency and compliance
7 GitHub stars
Targeted focus rubrics for Security, Performance, and Accessibility

ユースケース

Automated requirements verification for technical documentation
Pre-commit validation of security-critical code changes
Implementing multi-model quality gates in CI/CD pipelines

概要

主な機能

Structured JSON output with scores and blocking issues
Snapshot-based reproducibility using Git commit SHAs
Multi-model consensus for high-confidence pass/fail verdicts
Automated audit trails for transparency and compliance
7 GitHub stars
Targeted focus rubrics for Security, Performance, and Accessibility

ユースケース

Automated requirements verification for technical documentation
Pre-commit validation of security-critical code changes
Implementing multi-model quality gates in CI/CD pipelines