Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add social-media-skills/skills --skill thumbnail-designgit clone --depth 1 https://github.com/social-media-skills/skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/social-media-skills/skills/thumbnail-design)<a href="https://agentmods.dev/skills/social-media-skills/skills/thumbnail-design"><img src="https://agentmods.dev/badge/skills/social-media-skills/skills/thumbnail-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/social-media-skills/skills/thumbnail-design"><img src="https://agentmods.dev/badge/skills/social-media-skills/skills/thumbnail-design.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00223 | $0.01654 |
| Opus 5 | $0.00112 | $0.00827 |
| Sonnet 5 | $0.00045 | $0.00331 |
| Haiku 4.5 | $0.00022 | $0.00165 |
Grade A, and why
thumbnail-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
thumbnail-design
A design-craft skill — write the promise, design the concept + text, produce the brief, and plan the A/B test. The agent designs + briefs + plans; a human or image tool makes the raster; the YouTube thumbnail is uploaded + tested natively (no WoopSocial field).
The POV: a promise, not a poster — designed as a unit with the title
A thumbnail drives ~70–80% of click decisions, and the viewer decides in under a second — so its job is the click that leads to a watch. Two rules anchor everything: the thumbnail + title are one unit (the thumbnail shows, the title tells — they never repeat), and it's a promise, not a poster. That second rule is enforced by the platform: YouTube's native A/B test picks the winner by watch-time share, not raw CTR, so a misleading high-click thumbnail loses to an honest one. Design for mobile first — most viewing is on a phone, where simplicity and contrast win.
Read these first
- brand-profile — channel style/template (for consistency).
- the video and its title (often from youtube-publishing-and-metadata/youtube-long-form/ youtube-shorts).
The framework: CLICK
(Depth: references/the-click-framework.md.)
- C — Clarify the promise: one-sentence promise; thumbnail + title as a unit (show vs tell, never repeat); promise not poster.
- L — Lead with one focal point: one/two hero elements, minimal background, parseable in <1s; one directional element.
- I — Invoke genuine emotion: expressive genuine face + gaze if it adds context (skip it otherwise); match the expression to the content.
- C — Contrast + concise text: complementary high contrast (avoid YouTube red/white, 60-30-10, B&W test, light+dark); 0–4 bold sans-serif words + outline, not a title duplicate; mobile-legible at ~168×94; safe zones; 1280×720, 16:9.
- K — Keep testing: 3 meaningfully-different concepts, one variable; native Test & Compare picks by watch-time share (not Shorts, needs traffic); a human uploads + runs it; keep a log; brand-consistent template.
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 96 lines · 223 tokens per session scan A 1a9a4bf79eb4
thumbnail-design is a skill published in the GitHub repository social-media-skills/skills (77 stars, last pushed 8d ago), licensed MIT. It adds 223 tokens to every session and 1,654 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
smart-illustrator
An image-generation tool for adding illustrations to articles, creating slide presentations and infographics, or making cover images and thumbnails. It can also create prompts without generating the images.
xiaohongshu-cover
An AI-assisted cover design tool for Xiaohongshu notes, which are short posts on a Chinese social media platform. It studies visual elements from popular covers and produces design suggestions matched to the post's subject.
wechat-cover
An AI design tool for planning cover images for WeChat Official Account articles. It studies visual patterns from popular articles and uses them to suggest cover designs related to the article’s subject.
design-landing-page
Generate a deliverable landing-page HTML file (NOT a screenshot artifact). 1440px design width, mobile-responsive (CSS-only, no JS). Sections: hero (full-viewport) + 3 feature blocks + CTA. Designed to be opened in a browser, iterated on, and deployed as-is. Reads ./design/ /DESIGN.md, tokens.css, and components.html.…
design-system
Generate a brand-agnostic DESIGN.md (Google design.md spec format) plus tokens.css and showcase.html for any branding project. Reads ./design/ /DESIGN-PLAN.md if present, otherwise asks for an inline brief. Use AFTER design-plan and BEFORE design-ui-components. Triggers include "design system for [brand]", "build a…
design-carousel-slide
Generate a LinkedIn / Instagram multi-slide carousel — single HTML file with N square 1080×1080 .canvas divs, each representing one slide. Default N=5 (cover + 3 content + closing CTA). Cover and closing always use the brand's most distinctive treatments; middle slides are content-led. Reads ./design/ /DESIGN.md…