agentleFS
itqanlabitqanlab/agent-toolkit/.agents/skills/video-prompt/SKILL.md

video-prompt

Skill1 starsChanged 34 days agoMITView on GitHub

Write video prompts in the project's brand, sized and timed for each platform, ready for any video generator. Triggers: 'write a video prompt', 'prompt for a reel', 'make a prompt to animate this photo', 'prompt for a product video', 'give me a prompt I can paste into Kling', 'how long should this reel be', 'video prompt'. Reads the project's own brand file for palette, look, motion, tone and a don't list, and never stores brand values itself. Writes the prompt, a JSON with size, ratio, length and a suggested model, and on request short paste steps for a web tool. It never generates a video and never spends credits. Needs Node 18 or newer.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No brand file is fine: the prompt carries no brand style, and the tool says so.

    line 69
  2. Distinctive:

    Never invent brand values, and never copy them into this skill.

    line 70
  3. Distinctive:

    Read the brand file every time: Never write a brand value from memory.

    line 79

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video Prompt

Writes the prompt for a short video: in the project's brand, in the right size and length for the platform, with the motion spelled out and a don't list. It writes text only. It never generates a video, so it costs nothing.

The prompt then goes to a video generator. When the higgsfield-studio skill is installed, hand it over there: it quotes the price before anything is spent, and video is where credits go fastest. Otherwise the user pastes it into any web tool.

Who you are doing this for

Assume the person asking has never opened a terminal. You run the commands; they never do. Show them the prompt in plain words and say what happens next. If they are clearly comfortable with a terminal, let them drive.

Before anything else: check the dependencies

sh scripts/setup-deps.sh --check

Non-zero exit means Node is missing or older than 18. sh scripts/setup-deps.sh (Windows: .\scripts\setup-deps.ps1) prints the right install command for the machine and asks before running it; --yes skips the question.

Commands

node scripts/prompt.mjs --subject "..." [--type turntable] [--platform instagram_reel] [--variant square]
     [--size 1080x1920] [--seconds 7] [--reference] [--people] [--scene "..."] [--motion "..."]
     [--brand FILE] [--brand-mode full|minimal|none] [--paste "Tool name"] [--out DIR] [--force]
node scripts/prompt.mjs sizes        every platform, variant, size and accepted length
node scripts/prompt.mjs brand        which brand file is used and what was read from it
  • --subject is what the video shows, in plain words.
  • --type is the shot: turntable (default), hero-reveal, editorial, detail or animate-still. See references/prompt-craft.md.
  • --platform and --variant set size and default length from the table below. --size WxH sets the size directly.
  • --seconds sets the length. A length the platform does not accept is refused.
  • --reference when there is a photo of the real subject (image to video). animate-still always uses one.
  • --motion replaces the shot's motion line. The brand's own motion rules are still added after it.
  • --paste "Kling" also writes PASTE.md: short, general steps for pasting the prompt into that web tool.

Each run writes a folder with prompt.txt (the prompt alone, ready to copy) and prompt.json (the prompt, size, ratio, length, don't list, the brand file used, and a suggested Higgsfield model with ratio, clip length, number of clips and a cheaper draft model).

Platforms

PlatformVariantsSizeDefault length
instagram_reel, facebook_reel, tiktok, youtube_short1080x19207 to 10 s
instagram_story, facebook_story1080x19207 s
instagram_postportrait (default), square1080x1350, 1080x10807 s
facebook_feed, linkedin_postsquare (default), landscape1080x1080, 1280x720 or 1920x108010 s
youtube1920x108010 s
product_page1080x10805 s

node scripts/prompt.mjs sizes shows the exact length range each one accepts.

The brand

The brand comes from a Markdown file the project owns: --brand FILE, else BRAND_FILE, else brand.md, BRAND.md, brand/brand.md or docs/brand.md in the folder the command runs from. Its sections (palette, look, studio, motion, tone, don't) are found by their headings. The full contract and an example are in references/brand-file.md.

  • No brand file is fine: the prompt carries no brand style, and the tool says so. Offer to write one with the user. For video, also ask how things should move: pace, camera, cuts.
  • Never invent brand values, and never copy them into this skill. Read the file every time.
  • --brand-mode minimal keeps only the look and the don't list. none ignores the brand.

Handing it to Higgsfield

prompt.json has a higgsfield block: the model, a cheaper draft model, the aspect ratio, the length of each clip and how many clips. With the higgsfield-studio skill installed, use its create command with --prompt-file pointing at prompt.txt, the model, --aspect_ratio and --duration, and --image for a reference. It quotes the price first. Make the draft, check the motion, then render the final once.

Rules

  • Read the brand file every time. Never write a brand value from memory.
  • Short first. 5 to 10 seconds fails less and costs less. A longer piece is several clips.
  • One prompt, then look. Fix with one change at a time (references/prompt-craft.md lists the usual fixes).
  • Respect the log. Exit code 3 means a similar prompt was already written in this project. Show the user the earlier one. Use --force only when they want a new one.
  • Disclosure. Generated video counts as AI-generated material on most platforms. Tell the user to use each platform's AI label when they publish.

Updates

This skill is versioned. Its version is metadata.version in the header of this file. CHANGELOG.md in this folder lists what changed in each version, newest first.

To check for a newer version, open the address on the Latest: line of CHANGELOG.md and compare its top version with the installed one. If the newer one is ahead, read every entry between the two and tell the user what changed before anything is updated. A Breaking section means the user has to do something. To update, reinstall the skill from its source repository, the same way it was installed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.