---
title: Human-in-the-Loop Governance
created: 2026-05-15
updated: 2026-05-15
type: concept
tags: [open-scaffold, governance, workflow, public-boundary]
sources: [MISSION.md, ROADMAP.md, docs/SLICE_CLOSE_PROTOCOL.md]
confidence: high
contested: false
---

# Human-in-the-Loop Governance

Human-in-the-loop governance is the Open Scaffold boundary that agents and runtime harnesses can propose, execute, verify, and report, but humans keep final authority over taste, risk, publication, and scope pivots.

This matters because agentic development can move fast enough to blur responsibility. Open Scaffold makes the gate explicit: execution lanes return artifacts and evidence; the human owner or team decides whether the result is accepted, revised, published, or parked.

## Governance surfaces

- Plans define acceptance criteria and non-goals.
- Run packets define allowed scope and evidence obligations.
- PRs expose code review and merge gates.
- Release notes preserve accepted outcomes and known gaps.
- Operator surfaces can request approval, but they do not become durable truth.

## Good governance

A good slice states what an agent may do, what it may not do, and where human approval is required.

Related: [[evidence-first-development]], [[glass-cockpit]], [[source-of-truth-first-development]].