quorum

What This Skill Does

Quorum is an advanced multi-agent validation framework designed to ensure the quality, integrity, and consistency of AI-generated artifacts. It operates as a high-fidelity gatekeeper, sitting between initial generation and final production release. By utilizing a deterministic pre-screen stage followed by four independent, specialized AI critics, Quorum ensures that every document, code repository, or research report undergoes rigorous evaluation against predefined rubrics.

The framework is built on a philosophy of evidence-based judgment. Each critic is required to ground its findings in the actual text of the artifact, reducing hallucination and ensuring that critiques are actionable and objective. The deterministic pre-screen, which runs prior to any LLM-based analysis, performs ten critical checks, including searching for leaked credentials, PII (Personally Identifiable Information), syntax errors, broken hyperlinks, and unresolved TODOs, providing a vital first layer of security and hygiene.

Installation

To integrate this skill into your environment, use the OpenClaw CLI: clawhub install openclaw/skills/skills/dacervera/quorum

Ensure that you have your environment variables set for your chosen LLM providers (e.g., ANTHROPIC_API_KEY or OPENAI_API_KEY) and that you have initialized your local configuration using quorum config init to define your model tiers.

Use Cases

Quorum is designed for teams that require high-assurance automated output validation. Primary use cases include:

• Technical Documentation: Verifying that API references and architectural guides remain consistent with source code.
• Agent Configuration: Ensuring system prompts and YAML specifications adhere to safety standards.
• Research Synthesis: Validating long-form research papers against technical rigor rubrics.
• Code Quality Control: Enforcing hygiene standards across large codebases via automated batch scanning.

Example Prompts

1. "Quorum, validate the research paper at ./reports/q3_strategy.md using the research-synthesis rubric and output the verdict."
2. "Run a standard depth Quorum check on the current directory; specifically, ensure the agent configuration files are consistent with our baseline."
3. "Quorum, list all available rubrics and then perform a quick check on ./src/config.yaml."

Tips & Limitations

• Boundary Awareness: Always review portfolio/research-infrastructure/VALIDATOR-QUORUM-BOUNDARY.md before deploying. This document is mandatory for maintaining compliance with internal governance standards.
• Depth Selection: Use the quick profile for rapid iteration during development and reserve thorough for pre-release artifact finalization.
• Config Management: Ensure your quorum-config.yaml is kept updated as your preferred model capabilities change to maintain evaluation quality.
• Limitations: Quorum is a validation tool, not a creative tool. It assumes the provided rubrics are sufficient; ensure custom rubrics are well-defined to get the best results.