agentic-governance

You can enforce constitutional AI safety rules by applying the F1-F13 floor checks to validate reversibility, truthfulness, consensus, and clarity before any agent action is executed, ensuring safe operations.

Quad-Witness consensus is a BFT model used to validate AI actions. It requires a minimum consensus score of 0.75 to authorize operations, ensuring multiple independent verifications of safety compliance.

Implement a governance gatekeeper that verifies actions against 13 constitutional floors (F1-F13) and requires Quad-Witness consensus. This ensures all safety rules are met, involving human oversight if necessary for critical operations.

Yes, AI governance floor enforcement operates without external dependencies. The Skill functions independently using internal scripts and references to validate constitutional floors and consensus scores for AI actions.

Use constitutional AI governance when an AI agent performs critical operations like modifying system files or accessing sensitive data, requiring hardened floor checks and consensus validation to prevent harmful operations.

A 0.75 consensus score limitation means actions are authorized when 75% of witnesses agree. If consensus drops below this threshold, the AI action is blocked, requiring human oversight or protocol adjustment to proceed safely.