Agentic Covenants

A governance model for autonomous systems. Guardrails enforced by infrastructure are not bypassable; guardrails enforced by the AI tool are. Every claim below is backed by policy in the source repository, and every control names the bypasses it still has.

• Matrix18 cells

  If the agent decides to violate this concern, what stops it at this layer?

• Charter15 cells

  Who is allowed to put this agent into the world, under what authority, accountable to what policy, with what retirement criteria?

• Sentinels15 cells

  If this concern is breached, how do we know?

• Interventions15 cells

  How do I stop the bleeding now?

• Restorations15 cells

  How do I get back to a known-good state and not repeat this?

• Inventory15 cells

  What agents exist, what they touch, what threats they face?

Generated from peopleforrester/agentic-covenants at a7fdd66.