How does Council Gate handle disagreements between AI models?

It uses a multi-stage synthesis process to evaluate individual model responses, determining a final verdict and confidence score based on your configured threshold.

What happens if the AI consensus confidence is too low?

If the result falls below your confidence threshold or is marked as 'unclear', the skill returns an exit code that can pause the pipeline for human review.

Does Council Gate only check for security issues?

No, while it has a security focus option, you can customize the rubric to focus on performance, maintainability, or any other quality standard required for your project.

Can I use Council Gate with my existing GitHub Actions?

Yes, it is specifically designed to integrate into GitHub workflows using a dedicated job step and the llm-council-core package.

Where are the results of the quality checks stored?

All deliberations, including individual model responses and the final synthesis, are saved as JSON artifacts in the .council/logs/ directory for full auditing.

Council Gate

Name: Council Gate
Author: amiable-dev

byamiable-dev

•

Deployment & DevOps

Implements an automated AI quality gate for CI/CD pipelines using multi-model consensus to verify code and pull requests.

Council Gate leverages the power of the LLM Council to provide a robust, multi-model consensus layer for software development lifecycles. By running code snapshots through multiple AI models and synthesizing their results, it enables automated approval workflows and deployment gating based on customizable rubrics like security or performance. It integrates seamlessly with GitHub Actions, providing detailed audit transcripts and clear exit codes to control pipeline flow, ensuring that only code meeting high-confidence thresholds reaches production while flagging ambiguous cases for human review.

Key Features

01Seamless GitHub Actions integration for automated PR gating

02Support for 'unclear' verdicts to trigger manual human intervention

03Customizable quality rubrics and confidence thresholds

04Multi-model consensus for high-confidence quality verification

057 GitHub stars

06Detailed audit logs and transcripts for every deliberation

Use Cases

01Automating Pull Request approvals based on security and style rubrics

02Auditing AI-driven code reviews for compliance and quality standards

03Preventing deployment of code that fails multi-model verification checks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add amiable-dev/llm-council council-gate

For use in Claude.ai and ChatGPT

Download Skill