These principles constitute the constitutional core of the Agent Manifest v1.x series, beginning with v1.0.
They are stable and may only be modified under a major version change (e.g., v2.0).
Minor versions (v1.1, v1.2, etc.) may clarify definitions, refine language, or introduce optional profiles, but must not alter these core guarantees.
No autonomous agent should act without first declaring:
Every agent must declare a distinct and persistent Identity sufficient for reference and audit.
Every agent must declare a Responsible Party capable of answering for and stopping the agent.
Orphaned agents are structurally non-compliant for production deployment.
Every agent must declare:
An agent without explicit boundaries is structurally unbounded.
Every agent must declare a Risk Profile that includes known risks and uncertainty.
Unknown risk must be declared as unknown.
Every agent must declare a Stopping Authority specifying:
Agents without Stopping Authority are structurally uncontrollable.
Every agent must declare its Audit Surface, including what can be:
Opacity must be declared if present.
If an agent cannot truthfully declare the required structural elements defined above, the default structural position is non-authorization for autonomous operation.
These principles define the structural minimum for autonomous systems under Agent Manifest v1.x.