svenkatreddy/universal-video-skills / skills/video-prompt-craft
video-prompt-craft
VerifiedWrite text-to-video and image-to-video prompts that hold up in real generation. Shot-brief grammar, JSON prompting, negative prompts, and failure-mode hardening. Use when drafting, reviewing, or fixing a video generation prompt for any model.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.8 KB |
| references/failure-modes.md | markdown | 2.9 KB |
| references/i2v-prompting.md | markdown | 3.5 KB |
| references/json-prompting.md | markdown | 2.8 KB |
| references/negative-prompts.md | markdown | 2.0 KB |
| references/prompt-grammar.md | markdown | 3.4 KB |
references/json-prompting.md
View on GitHub# JSON Prompting
Prose prompts are easy to write and easy to silently break. Change one word in
a paragraph and nothing tells you what moved. JSON prompts fix that: every
decision has a named field, every revision is a diff, and pipelines can
generate or validate them mechanically.
This became the dominant serious-practitioner format in early 2026 for exactly
those reasons. Several models (notably Veo-class) also parse structured input
unusually well.
## The schema
```json
{
"duration_sec": 8,
"aspect_ratio": "16:9",
"subject": {
"description": "MARA, 34, sharp jaw, copper bob, olive bomber jacket",
"reference_id": "mara-01"
},
"setting": "night market in the rain, neon signs, canvas awnings",
"camera": {
"framing": "medium tracking shot, waist height",
"movement": "dolly keeping pace on subject's left, no shake"
},
"beats": [
{ "t": "0-3s", "action": "she weaves between stalls, neon signs sliding past" },
{ "t": "3-6s", "action": "she slows at a dumpling cart, steam washing over her face" },
{ "t": "6-8s", "action": "she looks up into the lens and smiles, barely" }
],
"audio": {
"dialogue": [{ "speaker": "vendor", "line": "...", "language": "Cantonese" }],
"foley": ["sizzle", "rain on canvas"],
"ambience": "night market chatter, distant traffic"
},
"preserve": ["Mara's face and jacket identical throughout", "three stalls on her right"],
"avoid": ["morphing faces", "extra limbs", "readable text on signs", "slow motion"]
}
```
## When JSON earns its keep
- **Multi-shot projects.** Generate the JSON per shot from one script and the
structure stays consistent where prose would drift.
- **Revision tracking.** `diff shot03-v2.json shot03-v3.json` shows exactly what
you changed. Try that with two paragraphs.
- **Pipelines.** Scripts can assemble, validate (required fields present?
beats sum to duration?), and translate these into per-model call formats.
- **Team handoff.** A JSON brief is unambiguous in a way prose never is.
## When prose is fine
Single exploratory shots. If you're just feeling out a look, write the
shot-brief in prose and move on. Convert to JSON once the shot is worth
keeping.
## Practical notes
- Keep field names stable across your project. The schema above is a starting
point. Adapt it, then freeze it.
- `reference_id` points at your asset registry
(`../../character-consistency/references/asset-registry.md`). The ID is the
contract; the description is the fallback.
- Validate beats against duration: the timed chunks should add up. A 30-second
script check catches more errors than a careful re-read.
- Some models accept JSON natively; others want it flattened to prose. The
model adapters (`../../model-adapters/`) note which is which. Either way,
author in JSON and render to prose when needed, never the reverse.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |