Consensus Guard Core

What This Skill Does

Consensus Guard Core is the foundational policy engine for the Consensus.Tools ecosystem. It acts as a shared library of deterministic guard primitives that ensure consistency across complex AI agent workflows. Rather than functioning as a standalone user-facing tool, it provides the underlying plumbing for hard-block taxonomies, weighted vote aggregation, reputation management, and idempotency control. By centralizing these logic-heavy operations, the skill prevents 'policy drift' where different agents in a swarm might apply conflicting rules to the same task. It implements strict schema enforcement for data artifacts, ensuring that every state change in an agent’s workflow is valid, replayable, and audit-ready.

Installation

The skill can be installed via the OpenClaw CLI using the command: clawhub install openclaw/skills/skills/kaicianflone/consensus-guard-core. Once installed, it integrates into your agent's runtime environment, typically using node or tsx. Developers should import the stable public API directly from the package root, avoiding deep internal paths to ensure compatibility with future updates.

Use Cases

• Multi-Agent Governance: Coordinating complex decision-making processes where multiple LLMs must agree on an action using weighted voting.
• State Integrity: Ensuring that agent state files (managed via CONSENSUS_STATE_FILE) remain uncorrupted through strict schema validation and idempotency keys.
• Reputation Tracking: Automatically calculating and updating agent reputation scores based on objective output quality and adherence to consensus policies.
• Audit Trails: Maintaining high-fidelity, indexed logs of board artifacts for retroactive analysis of agent decision-making paths.

Example Prompts

1. "Apply the consensus-guard-core schema validation to the current board artifact to ensure no unknown fields are present before proceeding with the task."
2. "Execute a weighted vote aggregation using the latest persona set defined in the consensus-guard-core configuration."
3. "Generate an idempotency key for the pending decision log to ensure this workflow can safely retry upon network failure without duplicating actions."

Tips & Limitations

• Policy Enforcement: This skill is strictly infrastructural. It does not perform tasks itself; it provides the rules that other skills must follow. Ensure your workflow triggers its validation methods before performing state-mutating operations.
• Environment Variables: Properly configure CONSENSUS_STATE_ROOT in your agent's environment, as all filesystem operations are relative to this root. Failure to define this will lead to errors in write-heavy workflows.
• Dependency Management: Because this package acts as a core dependency for other skills, treat it as a immutable infrastructure component. Do not manually mutate its internal state files unless you are performing an explicitly requested administrative audit.