---
name: h3-prompt-expert
description: "Create one structured reference-aware video execution brief from authorized inputs. Use when the distinct result and its preservation contract matter; unlike generate-video (relationship 0.68), choose this when structured reference-aware video execution brief is requested; choose generate-video for its narrower existing outcome. Do not use for generic media, source-provider tooling, automatic publication, or unverified capabilities."
---
# Video Prompt Expert

## Catalog

- Display name: Video Prompt Expert
- Text category: yes
- Categories: Text, Video
- Status: Ready
- Route label: Independent outcome contract
- Tone: purple
- Short description: Produce one structured reference-aware video execution brief with explicit inputs, boundaries, and acceptance checks.

## What this skill does

- Captures the specific requested outcome and authorized asset roles.
- Produces a reviewable plan before any paid media route.
- Hands only approved atomic work to currently verified native Atoms.

## How to use

Provide the desired result, authorized inputs, preservation requirements, ratio/duration when relevant, and acceptance criteria.

#### Example

```text
Create one structured reference-aware video execution brief from my authorized assets; preserve the stated facts and show the plan before media generation.
```

## Workflow

1. Confirm the user-visible outcome, authorized inputs, fixed facts, delivery format, and exclusions.
2. Produce a concise plan and wait when it materially controls paid downstream work.
3. Select the narrowest verified existing Atom for any media operation; use only its live schema and never hard-code source-provider models or tools.
4. Persist one durable operation key for each intended paid run, poll accepted receipts, and stop on outcome-unknown.
5. Review the requested preservation facts, output scope, ratio/duration, visible quality, and unwanted text or unsupported claims. Repair only the failing stage.

Read [source-provenance.md](references/source-provenance.md) for external-source provenance.

## User-facing output

- Media type: One structured reference-aware video execution brief
- Default quantity: 1
- Content per artifact: Approved plan, requested deliverable, and acceptance result
- Default layout: User-requested format
- Model policy: Existing verified native Atoms only
- Downstream use: Review, assembly, or publication after user approval