Coordinate cross-model deliberation across frontier models to produce a trusted answer.
npx skills add https://github.com/rachittshah/llmcouncil --skill council-rachittshah
Please help me install this Agent Skill. Skill: council Source: https://github.com/rachittshah/llmcouncil/tree/main Command: npx skills add https://github.com/rachittshah/llmcouncil --skill council-rachittshah
Orchestrates cross-model deliberation across frontier models to produce robust, trusted answers through structured protocols (voting, debate, synthesis, critique, red teaming, and verification).
Start the MCP server and invoke council_deliberate with your question to begin a multi-model deliberation.
High-intent search queries and answers about installing and using this skill.
Multi-LLM deliberation coordinates multiple frontier models to produce a single trusted answer through voting, debate, and synthesis. Use it for complex questions requiring consensus, verification, and risk assessment.
Start the MCP server and invoke the council_deliberate function with your question to trigger a multi-model deliberation run. The protocol-driven orchestration handles model coordination and chairman synthesis automatically.
Yes, the council includes a cost estimation tool to project expenses before executing deliberation runs. You can also adjust default configurations and check status to control multi-LLM orchestration spending.
MAV verification cross-checks candidate answers with multiple models to validate results. You can also apply critique and red teaming protocols for peer review and adversarial testing to surface weaknesses.
Debate protocols support multi-round argumentation with optional KS adaptive stopping to converge on a stable view. A chairman synthesizer model then merges the parallel results into a single authoritative answer.
Avoid multi-model deliberation for simple queries where single-model responses suffice, as cross-model consensus, voting, and critique protocols introduce higher latency and API costs from orchestrating multiple frontier models.
mahmoud20138Configure Claude Code plugins to access external services via MCP servers.
luongldptitGenerate a four-voice adversarial verdict with explicit dissent for ambiguous decisions.
Marcus-Mok-GHExecute multi-agent voting with MAKER-style error correction for consensus decisions.
RyanDev1stReview engineering proposals through cross-model adversarial analysis and generate failure hypothesis reports.
