v0.2.0 to v0.3.0

---

name: narova-3d-production

description: >

Use this optional Narova companion when creating, directing, or repairing an

authored 3D video whose principal subjects, world, asset path, spatial design,

materials, lighting, camera, interaction, or inspection need specialist

judgment. Use for abstract or realistic 3D, recognizable or articulated

subjects, layered environments, mechanisms, camera-led scenes, and mixed

2D/3D where authored 3D has an intentional role. Do not use it merely because

a request says animation, cinematic, explainer, product video, Three.js, or

render; ordinary 2D motion, browser walkthroughs, AI-generated clips, and

direct render commands remain core Narova work.

license: Apache-2.0

metadata:

author: ammar-hasan

---

# Narova 3D Production

Add a small expert lens to authored 3D work. Keep the `narova` skill in charge

## Direct lightly

1. **Give 3D a job.** State what authored 3D lets the viewer perceive that a

simpler or different medium would not. If it adds no material value, use the

better Narova path and stop using this companion.

2. **Choose what the audience will receive.** Decide what the viewer should

notice, understand, or feel, then choose intentional final forms for the

principal subjects and world. A primitive, model, captured or generated

element, and mixed composite are all valid. Do not silently turn a blockout

or cheapest available shape into the final representation.

3. **Route the material risk.** Choose at most the one or two areas most likely

to weaken the idea and read only their references:

- read [subjects and assets](references/subjects-and-assets.md) when final

representation, acquisition/generation, model readiness, rigging,

animation, rights, or a missing production capability matters;

- read [scene direction](references/scene-direction.md) when space,

appearance, staging, camera, visibility, interaction, or compositing is

uncertain; and

- read [inspection](references/inspection.md) when the visual premise,

movement, structured state, evidence limits, or a stop decision matters.

A straightforward scene may need none. Do not perform a full-manual pass.

4. **Prove before scaling.** For risky work, show one delivery-scale visual

target without its rationale. If motion or interaction carries the idea,

next show a short sample using that accepted representation. Expand only a

premise that reads from the artifact itself.

5. **Finish through core Narova.** Use available specialist capabilities only

when the chosen form needs them, then hand assets, deterministic motion,

rendering, review, release, and delivery to the ordinary Narova lifecycle.

## Keep confidence honest

Connect decisions to what was rendered or otherwise inspected. Keep factual

checks separate from perceptual judgment. Do not give the perceptual reviewer

the author's plan, symbol mapping, or run report. Treat an automated visual

critic as advice until it has agreed with independent owner judgments on this

kind of work. Render success does not prove recognition, collision, physical

simulation, target visibility, taste, or hidden scene state.

If the chosen form needs sourcing, generation, refinement, rigging, animation,

simulation, or inspection that no available capability supplies, change the

form with the user, name the gap, or stop. Never pretend this skill's prose

performed that operation.

## Preserve creative range

Do not default the work to a palette, realism level, low-poly language, material

system, lighting rig, camera move, shot list, density, pacing, duration,

narration, caption style, transition, platform, or aspect ratio. Minimal and

rich scenes, graphic abstraction and articulated realism are all valid when

intentional. Use hard constraints only where user intent, compatibility,

evidence, or visible correctness would otherwise break.