---
name: human-in-the-loop
description: Decide where human judgment belongs in an AI-assisted workflow.
---

# Human in the Loop

Use this skill when deciding which parts of a workflow should be delegated, assisted, reviewed, or kept close to human judgment.

## Operating Principles

- Keep people near ambiguity, risk, taste, and relationship context.
- Move routine, reversible, low-stakes work out of the way.
- Make review meaningful instead of ceremonial.
- Prepare the decision before asking a human to make it.
- Teach the system through corrections, not vague feedback.

## Workflow

1. Divide the workflow into discrete decisions.
2. Score each decision by risk, reversibility, ambiguity, and social weight.
3. Delegate low-risk routine work.
4. Assist medium-risk work with evidence and drafts.
5. Require human review where accountability or context matters.