agentleFS
mrdainamimrdainami/omni-video-director/.claude/skills/graphic-design/SKILL.md

graphic-design

Skill16 starsChanged 4 months agoMITView on GitHub

Generate an on-screen graphic (infographic, stat card, title, logo lockup) with GPT-Image-2 on kie — up to 16 references, real legible text. Use for type=graphic beats after the brief is crafted and the user approves the spend.

6 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use typography and ui-design.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • soilmassSkill

    typography

    Typography and kinetic text animations including character reveals, typewriter effects, scramble/decode, wave motion, and animated gradients. Use when creating text reveals, headline animations, number counters, or any text-based motion effects.

    Also in 3 projects that use this

  • SentrySkill

    ui-design

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    No refs? use gpt-image-2-text-to-image (drop input_urls).

    line 33
  2. Distinctive:

    Every word quoted verbatim in the TEXT section with placement + colour.

    line 40

“Common” is measured: kept by at least 6 in 10 of the 6 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

graphic-design — GPT-Image-2 stills (kie)

Paths — relative to the active project projects/<slug>/. Every graphic this skill makes lands in assets/refs/ — whether it's fed into Omni as a reference or overlaid directly on a type: graphic beat. Name it for the beat (e.g. assets/refs/beat3-statcard.png).

For anything with real text or exact logos — the things Omni garbles. Model: gpt-image-2-image-to-image (or -text-to-image with no refs).

Before spending — the gate

Billed on submit. Show the estimate, get a go. (GPT-Image-2 ≈ a few credits per still; the exact figure prints as creditsConsumed.)

Inputs (from the beat)

  • The role-map brief from craft-prompt.
  • Reference images (logos as real transparent PNGs, product shots, style refs), ≤16.
  • Aspect = the input video's aspect.

Build the body + run

  1. Host each reference: URL=$(bash scripts/kie.sh upload <ref.png>).
  2. Write assets/refs/<name>.json:
{ "model":"gpt-image-2-image-to-image",
  "input":{
    "prompt":"=== REFERENCE ROLE MAPPING === ...\n=== TEXT RENDERING CONSTRAINTS === ...",
    "input_urls":["<ref url>","<ref url>"],
    "aspect_ratio":"16:9",
    "resolution":"2K" } }
  • input_urls: up to 16 references (this is why we use GPT-Image-2 — many refs).
  • resolution: 1K iterate · 2K default · 4K hero (1:1 can't 4K).
  • No refs? use gpt-image-2-text-to-image (drop input_urls).
  1. TID=$(bash scripts/kie.sh submit assets/refs/<name>.json) → wait → download to assets/refs/<name>.png.
    • GPT-Image-2's kie endpoint sometimes throws a transient 500 — just resubmit (a failed job is 0 credits). 4K queues slower; poll patiently.
    • How to wait (mandatory): submit once, then run kie.sh wait in the background as the single poller. Do NOT hand-roll a recordInfo loop or post a message on every poll, and never state elapsed time you didn't actually measure. One "submitted, waiting" line, then silence until it resolves or errors (the harness re-invokes you when the background wait finishes). A live taskId is already billed — never resubmit it.
  2. Record creditsConsumed. Status → 👀 review.

Text + logos (non-negotiable)

  • Every word quoted verbatim in the TEXT section with placement + colour.
  • Logos passed as numbered images, "use Image N exactly, do not redraw" — never model-drawn.

Then

GATE 2 — user reviews assets/refs/<name>.png. On approval, assemble overlays it at the beat's timestamp (case B).

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.