consensus

Consensus from multiple AI models works by querying Claude, Gemini, and other providers simultaneously, then aggregating their responses using configurable strategies like synthesis, majority voting, or weighted scoring to reduce bias and improve decision confidence.

Yes, parallel multi-model querying enables cross-validation across providers. You invoke Claude, Gemini, and OpenAI together, apply error handling per provider, and synthesize responses to verify facts and catch single-model hallucinations.

Five consensus strategies are supported: all_responses returns individual outputs, synthesize merges insights into one answer, majority uses vote-based selection, weighted applies custom scoring, and first_valid returns the earliest successful response with graceful degradation.

Per-provider timeouts and error handling enable graceful degradation; if one model fails or times out, the orchestration continues with remaining providers and applies your chosen synthesis strategy to available responses.

No, the Skill coordinates whichever providers you configure. You can use any combination of anthropic-cli, google-generativeai, and openai-cli; parallel orchestration and strategy-driven aggregation work with your selected subset.

Query multiple models in parallel on the same question—for example, microservices vs. monolith trade-offs—then synthesize their diverse recommendations to weigh trade-offs from different perspectives before committing to a high-stakes architectural decision.