Higgsfield YouTube Thumbnail Skill for Claude: Truthful, High-Impact Covers

Makes Claude create YouTube thumbnails and vertical video covers through Higgsfield, with face references and edits.

Installs: 110.1K; on skills.sh
Stars: 1.8K; on GitHub
Files: 3; besides SKILL.md
Updated: Oct 2026; last repo commit

higgsfield-youtube-thumbnail has Claude brainstorm truthful thumbnail concepts, render them with the higgsfield CLI, check each result and make only the edits you ask for. It keeps up to three referenced people's identities, can use a logo flat or as a 3D object, and renders in 16:9, 9:16 or 4:5.

It triggers on requests like "make a YouTube thumbnail". It delivers a clean image with no text by default and offers a text overlay with five presets on request. It needs the CLI signed in, and SKILL.md says this workflow runs on credits. Generations are capped at 16 per job. Two reference files hold 16 concept frameworks and the overlay recipes.

Good for

YouTubers and Shorts creators
Marketers making video covers for Instagram and social
Teams who want several variants to pick from

Install higgsfield-youtube-thumbnail

Claude Code

Copy this prompt.

Install the higgsfield-youtube-thumbnail skill in Claude Code. Run this command:

npx skills add higgsfield-ai/skills

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

The README says each install method also handles the Higgsfield CLI install and sign-in as part of skill setup. It lists gh skill install (GitHub CLI 2.90 or newer) and a setup script as other routes. All routes take the whole repo; the README lists 9 skills.

From the project README. Source.

What's in the skill

When it loads

Create high-click-through YouTube thumbnails and vertical video covers through the Higgsfield CLI. Builds a truthful information-gap concept, preserves up to three referenced identities, supports logos and controlled variants, renders the main image with Nano Banana Pro, and applies focused Seedream edits. Use when: "make a YouTube thumbnail", "thumbnail for this video", "MrBeast-style cover", "Shorts cover", or "Instagram video cover". Chain after any video workflow once its truthful topic and visual direction are known. NOT for producing the video itself (use higgsfield-generate), product catalog photos (use higgsfield-product-photoshoot), or marketplace cards (use higgsfield-marketplace-cards).

Sections in SKILL.md

Bootstrap
UX rules
Intake gates
Concept gate
Reference order
Prompt contract
Split layouts
Optional 3D logo
Main render
Post-render gate
Surgical tweaks
Text overlay
Delivery
Reference files

Key facts

Author: Higgsfield AI (higgsfield-ai)
Repository: higgsfield-ai/skills
Installs on skills.sh: 110,065 (crawled 2026-11-02)
GitHub stars: 1,843 (whole repo)
License: MIT
Last repo commit: 2026-10-27
Files in the skill folder: 4 files: SKILL.md, agents/openai.yaml and 2 reference files. No scripts.
skills.sh audits: Gen Agent Trust Hub: fail, Socket: pass, Snyk: pass
Works in (per README): Claude Code, Cursor, Codex and other agents that load Markdown-based skills

What it includes

Installs the higgsfield CLI if missing, checks sign-in and the model catalog
Runs code · Bootstrap
Collects topic, truthful promise, people, logo, optional headline and ratio
Intake gates
Brainstorms at least five truthful concepts across 16 frameworks, picks the strongest
Concept gate
Renders in 4K with nano_banana_pro, passing face and logo references
Runs code · Main render
Renders the logo as a 1:1 3D object first with gpt_image_2
Runs code · Optional 3D logo
Checks identity, stray text and readability at 120px; retries a hard failure twice
Post-render gate
One edit at a time with seedream_v5_pro: expression, background or rim light
Runs code · Surgical tweaks
Five presets: Beast, Fire, Neon Lime, Clean Glass and Marker
Text overlay

From the official docs

Permissions

Medium risk

It uploads your face photos and logos to Higgsfield and spends your credits on every render. Set the number of variants before it starts, and only share photos you may use.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Reference photos and logos go to Higgsfield
Local paths passed with --image are auto-uploaded. Share only photos of people who agreed to appear.
It installs a CLI from the internet if missing
If `higgsfield` is not on your PATH, SKILL.md has Claude run the official install script from raw.githubusercontent.com. Install the CLI yourself first if you want to read it.
Every render uses credits
SKILL.md says the workflow runs on credits and caps a job at 16 generations. Ask for one concept if you want to limit spend.
Every install route takes the whole repo
npx skills add higgsfield-ai/skills, gh skill install and the Claude Code plugin all install every skill in the repo, and the README says setup also installs the Higgsfield CLI and signs you in. There is no single-skill command in the README.
It writes temporary files and may download a render
SKILL.md has Claude write the final prompt to a temporary text file and pipe it to the CLI. If an edit is rejected by older CLI versions, it says to download the picked result_url to a local image and retry.

Technical details

Talks to	Higgsfield (renders, uploads). raw.githubusercontent.com (CLI installer, if missing)
Sign-in and keys	Sign-in is done by you with `higgsfield auth login`. SKILL.md does not say where the CLI keeps the session.
Scopes	Your Higgsfield account (credits)

Questions

What does higgsfield-youtube-thumbnail do?

It makes Claude create YouTube thumbnails and vertical video covers through Higgsfield. It plans a truthful concept, renders it with your face and logo references, checks the result and makes only the edits you request.

How do I install it?

Use the install box at the top of this page. The README says each install method also handles the Higgsfield CLI install and sign-in as part of skill setup, and lists npx skills, gh skill install, a Claude Code marketplace and a setup script.

Does it cost credits?

Yes. SKILL.md says this workflow must run on credits in the CLI. It caps a job at 16 total generations, and for a variant set it offers about four.

Will it put text on the thumbnail?

Not by default. It delivers a clean image with no text. When you ask for a headline overlay, it uses one of five presets from a reference file, and bakes text into the image only if you ask for that.

Does it use my face?

Only if you provide photos. If a concept needs a person and no face photo is given, SKILL.md says Claude must ask whether to use you, another provided person or a generic character, and never choose silently.

Is it safe?

It bundles no scripts. It runs the Higgsfield CLI, may install it with a curl script, and uploads your reference photos and logos to Higgsfield. SKILL.md says never to invent claims or statistics that are not true of the video.

Information

updated Nov 2, 2026

Repository: higgsfield-ai/skills
Folder: higgsfield-youtube-thumbnail

higgsfield-ai/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

higgsfield-ai/skills README · github.com
INSTALL.md · github.com
SKILL.md · github.com