How does the Council handle API rate limits?

The skill includes a built-in retry strategy with exponential backoff and can stagger model launches to avoid hitting 429 errors from external providers.

What models does the Council skill support?

The Council skill supports external models including Gemini, Codex, Qwen, GLM-5, and Kimi via their respective CLIs, in addition to internal Claude Opus and Sonnet subagents.

What is 'Blind Mode' in the Council skill?

Blind mode strips Claude subagents of their native tool access (Read, Grep, Bash), putting them on equal footing with external models to compare tool-assisted findings against raw analysis.

Is it safe to send my code to external models?

Yes, the skill includes a pre-flight secret scanning gate using gitleaks and uses XML delimiters to prevent prompt injection, ensuring a more secure review process.

AI Consensus Council

Name: AI Consensus Council
Author: rube-de

byrube-de

•

협업 도구

Orchestrates multiple external AI models to provide consensus-driven code reviews, architectural planning, and adversarial analysis.

The Council skill empowers Claude with a 'jury of its peers' by invoking a diverse array of external LLMs—including Gemini, Codex, Qwen, GLM-5, and Kimi—alongside specialized Claude subagents. It is designed for high-stakes decision-making, providing a multi-layered approach to code quality, security audits, and architectural strategy. By aggregating perspectives from different model families, it minimizes individual model bias and delivers a weighted, scored synthesis of findings. The skill includes sophisticated orchestration features like automated rate-limit handling, staggered model launches, and a 'blind mode' for unbiased comparisons, ensuring that every piece of feedback is rigorous and reliable.

주요 기능

01Specialized execution modes: adversarial, planning, consensus, and quick-scan

02Built-in security hardening including secret scanning and prompt injection prevention

03Multi-model consensus from Gemini, Codex, Qwen, GLM-5, and Kimi via CLI

045 GitHub stars

05Dual-layer architecture combining external model diversity with deep Claude subagent analysis

06Automated rate-limit management with exponential backoff and staggered launches

사용 사례

01Identifying subtle edge-case bugs and performance bottlenecks using adversarial AI perspectives

02Performing comprehensive security and architectural reviews before critical production deployments

03Resolving complex technical debates through multi-agent consensus and weighted scoring

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rube-de/cc-skills council

For use in Claude.ai and ChatGPT

Download Skill