← All work

SELECTED WORK / GOVERNOR AGENT

Decision before action.

A governance control for AI-assisted software work: it makes authority, policy, permits, scope, and evidence explicit before a decision is made.

AUTHORITY→POLICY→DECISION→EVIDENCE

01 / SCOPE

The problem

A change can pass tests and still violate ownership, authority, architectural policy, or permitted scope. Governor addresses that gap between being able to execute and being authorized to do so.

02 / GOVERNANCE DECISION

Three explicit outcomes

01

ALLOW

Authority, permitted scope, and required evidence are present.

02

DENY

A required condition is missing, or the work falls outside applicable policy or permit.

03

ESCALATE

The decision may be legitimate, but it exceeds the agent's authority and returns to a person with evidence.

03 / AUTHORITY BOUNDARY

Authority remains human

Deterministic rules govern the automatic outcome within their boundaries. The advisory component can gather and interpret evidence, but cannot override policy; ambiguous or higher-authority decisions escalate to a person.

04 / AUDITABILITY

CONTRACTS+VALIDATORS+AUDIT RECORD+DIGEST

Decisions are recorded as append-only audit records with controlled validator execution and digest verification. Secret values do not enter prompts or audit records, and raw validator output stays outside the record.

05 / PUBLIC BOUNDARY

Public status

The public demonstration uses a synthetic factory and reproducible ALLOW, DENY, and ESCALATE scenarios. It does not operate on GoNucleo's private factory and has no authority to mutate real projects.

06 / PUBLIC EVIDENCE

Available public evidence

The public repository contains the synthetic scenarios, deterministic rules, and architecture and security documentation for the demonstration.

GOVERNOR AGENT / REPOSITORY ↗

EXTERNAL PUBLICATIONS

External publications

Technical articles published on AWS Builder Center develop Governor Agent's architectural decisions in public.

  1. AWS Builder Center
    Building Governor Agent — Keeping Human Authority Outside the Model
  2. AWS Builder Center
    Designing Deterministic Governance Gates Around AI Agents
  3. AWS Builder Center
    Designing Evidence Boundaries for Governed AI Agents