agentleFS
Sign inStart free
mehbylabsmehbylabs/creator-skills/skills/creator-thumbnail/SKILL.md

creator-thumbnail

Skill7 starsChanged 16 days agoMITView on GitHub

Design high-click YouTube thumbnails and vertical covers (Shorts, Reels, TikTok) with whatever image provider the user has, compositing headline text locally so it is always crisp. Plans the concept, generates 2–3 variants, keeps faces and logos faithful via reference images, and checks readability at small size. Use when the user asks for a YouTube thumbnail, video thumbnail, Shorts/Reels cover, podcast episode cover, or "make my thumbnail pop". Not for general images (creator-image) or channel branding (creator-brandkit).

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    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. Distinctive:

    Ask for no text in the image.

    line 37
  2. Very specific:

    Put the text in the space reserved in Step 2 (--align left for left-third space) and never over the face/subject.

    line 52
  3. Distinctive:

    Check the bottom-right corner (YouTube's timestamp covers it) is free of key content.

    line 60
  4. Distinctive:

    Don't put real people who aren't the user into the thumbnail without their consent.

    line 66

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Creator Thumbnail

Step 1 — Concept (before generating anything)

From the title/topic, write one concept in three lines and confirm only if the user seems to care about direction; otherwise proceed:

  • Promise — the single idea the viewer must get in 1 second (truthful to the video).
  • Focal subject — one face/object, big, high contrast. Faces: strong, honest emotion.
  • Headline — 0–4 words that add to the title, never repeat it. Optional.

Rules in references/thumbnail-rules.md (composition, contrast, text, honesty).

Step 2 — Generate the art (no text)

  • Ratio 16:9 (YouTube) or 9:16 (vertical cover). Size 2k if a provider allows it.
  • Reserve space for the headline: generous empty space on the left third for a headline.
  • User's face/logo/product → pass photos with --ref and say what each is for.
  • Ask for no text in the image; the headline goes on in Step 3.
python3 scripts/creator.py image --prompt "<art prompt>" --ratio 16:9 --size 2k --out ./out/thumb-a.png [--ref face.jpg]

Make 2–3 variants that differ in one big way (emotion, color, or composition), not tiny tweaks. Pin the provider after the first success.

Step 3 — Headline

python3 scripts/media.py text --image ./out/thumb-a.png --out ./out/thumb-a-final.png --text "TWO\nLINES" --pos center --align left --size 0.16 [--box "black@0.55"] [--font <brand heading font>]

Put the text in the space reserved in Step 2 (--align left for left-third space) and never over the face/subject. Use the brand font/colors from config brand if present. Max ~4 words; break into two short lines with \n when it's wide.

Step 4 — Check

  1. Downscale preview: ffmpeg -y -i final.png -vf scale=320:-1 small.png, then look at it — subject and words must read at that size.
  2. Check the bottom-right corner (YouTube's timestamp covers it) is free of key content.
  3. Present variants side by side with one line each on what differs; recommend one.

Rules

  • Truthful: no fake quotes, fake screenshots, or implied events that aren't in the video.
  • Don't put real people who aren't the user into the thumbnail without their consent.
  • 1280×720 minimum for YouTube; under 2 MB (ffmpeg -i in.png -q:v 3 out.jpg if needed).

References

  • references/thumbnail-rules.md — composition, text, color, examples
  • references/image-prompting.md, references/media-ops.md, references/providers.md

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.

Check your setup, free