Back to directory

Council of High Intelligence

A structured multi-perspective deliberation tool for hard decisions, assigning a question to independent analytical personas that must disagree directly before returning a verdict with dissent, kill criteria and a next action.

Council of High Intelligence

Council of High Intelligence runs structured, multi-perspective deliberation for decisions across Claude Code, Codex, Gemini CLI and OpenCode. It assigns a hard question to deliberately different analytical personas, keeps their first positions independent, forces direct disagreement, and returns a verdict that preserves unresolved questions, dissent, kill criteria and the next concrete action, distinguishing facts from inference, assumptions and unknowns along the way.

A full council runs independent analysis, cross-examination and a final synthesis for high-stakes decisions with competing frames; a quicker panel restates and analyzes more narrowly when breadth matters more than a full adversarial round; and a two-member duo debate suits a decision defined by one central tension. The project recommends against convening a council merely to manufacture support for an answer already decided, and against using it for factual lookups or cheap, reversible experiments.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.