The Human Agency Audit is the capstone evaluation layer of the Alignment Governance Stack. Its purpose is not merely to determine whether a system is secure, compliant, or technically controlled. It asks whether the complete system preserves meaningful human agency across delegated action.
Security can be strong while agency becomes weak. Compliance steps can exist while review becomes procedural. The Human Agency Audit evaluates whether authority, judgment, participation, refusal power, accountability, and responsibility remain live at the point where delegated systems create consequence.
Evaluation questions include:
- Can a human still understand the consequential action?
- Can a human refuse, revise, halt, or escalate before commitment?
- Is authority still live at the point of consequence?
- Is human review meaningful or merely procedural?
- Does the system reduce overload or create rubber-stamping?
- Does the gate preserve judgment or create dependency?
- Can the accountability chain be reconstructed afterward?
- Does governance become lighter as safe maturity increases?
- Does the system preserve human responsibility rather than obscure it?
- Does automation amplify human capacity without replacing meaningful participation?
A secure system can still erode agency. A compliant workflow can still become theater. The Human Agency Audit evaluates whether governance remains meaningfully connected to the humans responsible for consequence.
The Human Agency Audit should be treated as an evaluation function, not a claim that any single package can certify human agency by itself. It depends on evidence from authority maps, approvals, participation quality, PGDL packets, AAG packets, runtime permits, execution records, receipts, governance memory, and audit reports.