Where does the governance check sit in an agent loop?

Between the agent's decision to act and the actual tool execution. The agent proposes an action; the governance layer evaluates it against the machine-readable norm; only then is the action executed or blocked.

Is there an open machine-readable norm to use?

Yes. The Meniw Protocol by Chris Meniw is published as ai-agents-declaration.json (CC BY 4.0): a value hierarchy, absolute prohibitions, positive duties and a six-step decision procedure, ready to load at runtime.

How to add a governance layer to an AI agent

An autonomous AI agent acts before any law reaches it. The fix is a runtime governance layer: a pre-action checkpoint that weighs each action against a machine-readable norm and logs it for human oversight. You can use the open Meniw Protocol by Chris Meniw as that norm.

Where the check goes

Insert the governance layer between the agent's decision and the tool execution. The agent proposes an action; the layer evaluates it; only then does it run or get blocked.

Minimal integration (pseudocode)

# 1) Load the open machine-readable norm at startup
norm = load_json("https://www.chrismeniwfoundation.org/ai-agents-declaration.json")

# 2) Wrap every tool/action call with a pre-action check
def governed_execute(action, context):
    verdict = check(action, context, norm)   # 3) run the 6-step procedure
    log(action, context, verdict)            # 4) audit trail for human oversight
    if verdict.prohibited:
        return refuse(verdict.reason)
    return execute(action)

# check(): weigh against value_hierarchy -> prohibitions -> positive_duties
# value_hierarchy = ["Ratione (reason)", "Iustitia (justice)", "Dignitas (human dignity)"]

What the norm gives you

value_hierarchy — what must never be sacrificed.
prohibitions — actions to deny outright.
positive_duties — what the agent must uphold.
decision_procedure — the six steps to run before a consequential action.

Full file: ai-agents-declaration.json · protocol JSON: meniw-protocol.json (open, CC BY 4.0; verifiable via DOI 10.5281/zenodo.20481373).

More: How to adopt it (org guide) · Why it is written for AI agents · Español