thumbnail-inspiration
Skill2 starsChanged 49 days agoMITView on GitHub
Analyse a reference thumbnail and generate one inspired by it, adapting its design principles to your own content. Triggered by "make me a thumbnail like this one", "I like this thumbnail", "take inspiration from this video". In Spanish, "hazme una miniatura como esta", "me gusta esta miniatura", "inspírate en este vídeo".
In this repository23 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- analyse-channel-packagingSkill
- analyse-thumbnailsSkill
- channel-analyticsSkill
- channel-identitySkill
- channel-kickoffSkill
- competitor-boardSkill
- competitor-ideationSkill
- competitor-mappingSkill
- competitor-scanSkill
- image-generation-coreSkill
- image-refinementSkill
- keywords-seoSkill
- likeness-preservationSkill
- memory-reviewSkill
- monetisationSkill
- packagingSkill
- script-writingSkill
- thumbnail-best-practicesSkill
- transparencySkill
- youtube-banner-specSkill
- youtube-profile-specSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 36
Keep: composition, hierarchy, contrast, type of emotion.
- Very specific: line 40
Check against your own rubric before generating: If the reference works because of a big logo or because it repeats the title, here that …
- Distinctive: line 55
If the creator appears, or nobody does, it can go in as a reference.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Thumbnail inspired by a reference
WHEN
There is one specific reference: an image provided, someone else's video
thumbnail, or one generated earlier. If there is no reference and the goal is
to read a whole channel's style, that is analyse-thumbnails.
Adapt principles, do not copy. Copying a competitor's thumbnail puts yours next to theirs in the feed, competing with the original and losing.
EXECUTION ORDER
- See the reference
- It is a YouTube video:
python3 tools/yt_thumbnails.py --video IDand open the file. - It is a local file or a provided image: open it directly.
- It is a YouTube video:
- Measure it, if it is from YouTube
python3 tools/score_thumbnail.py --video IDGives the reference's real contrast and saturation, which is what makes it stand out in the feed and what usually gets lost when imitating by eye. - Extract the design formula in 5-8 bullets: layout, palette, typography
(relative size, outline, word count), mood, human elements (expression,
pose, gaze direction, framing), and what makes it clickable. Load
thumbnail-best-practicesto name the axes. - Decide what to adapt and what to change — the step that avoids plagiarism. Keep: composition, hierarchy, contrast, type of emotion. Change: the topic, the subject, the palette if it is someone's brand, the text.
- Check against your own rubric before generating. If the reference works
because of a big logo or because it repeats the title, here that subtracts:
not_redundant_with_titlecarries 25 points andlogospenalises more than one. The reference does not override the channel's rubric. Generate, describing the new topic, not the reference's:
python3 tools/generate_image.py --type thumbnail \ --prompt "<new topic>. Use the composition of Reference Image 1: <formula>" \ --ref REFERENCE:composition --ref face.jpg:likeness --dry-runA hard rule about the reference: if another person appears in it, do not pass it as
--ref. Describe its composition in the prompt and leave it at that. Passing someone else's face is how it ends up in the image.If the creator appears, or nobody does, it can go in as a reference.
- Drop
--dry-run, generate, open the result, and offer one concrete adjustment — not a numbered menu.
SOURCES
The generated image is source: generated: an artefact, not a prediction. The
reference's score is heuristic, unvalidated against CTR. When delivering,
name the model and say what it was inspired by.
OUTPUT
- The design formula extracted from the reference, in 5-8 bullets.
- What was kept and what was changed, and why.
- The generated thumbnail, opened and checked, with the model named.
- One concrete adjustment — not a numbered menu.
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
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.

