baoyu-cover-image skill: have Claude generate a cover image for your article

Makes Claude generate a cover image for your article, tuned by type, palette, rendering, text and mood.

Installs: 32.3K; on skills.sh
Stars: 27.2K; on GitHub
Files: 34; besides SKILL.md
Updated: Nov 2026; last repo commit

baoyu-cover-image is a skill from Baoyu's skills repo. Claude reads your article, recommends a cover type, color palette, rendering style, text level and mood, and asks you to confirm. It then saves the image prompt and generates a cover at 16:9 by default, or another ratio such as 2.35:1.

The first run asks a few preference questions and saves them. The skill does not draw by itself: it needs an image backend, such as baoyu-image-gen with a provider key in Claude Code. Generation uses your provider credits.

Good for

Bloggers and newsletter writers who need a cover for each post
Content teams keeping a consistent cover style
WeChat article authors; the README pairs it with the article and publishing skills

Install baoyu-cover-image

Claude Code

Copy this prompt.

Install the baoyu-cover-image skill in Claude Code. Run this command:

npx skills add jimliu/baoyu-skills --skill baoyu-cover-image --skill baoyu-image-gen

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, An image provider API key.

baoyu-cover-image does not draw images itself. It hands the saved prompt to an image backend, and the SKILL.md names baoyu-image-gen as an example backend, so the command also installs it from the same repo. The README lists more providers (Azure OpenAI, Z.AI, MiniMax, Seedream, Replicate and others). If the Codex CLI is installed and logged in, baoyu-image-gen can use it instead of a key. The README advises installing only the skills you need.

From the vendor's docs. Source.

What's in the skill

When it loads

Generates article cover images with 5 dimensions (type, palette, rendering, text, mood) combining 11 color palettes and 7 rendering styles. Supports cinematic (2.35:1), widescreen (16:9), and square (1:1) aspects. Use when user asks to "generate cover image", "create article cover", or "make cover".

Sections in SKILL.md

User Input Tools
Image Generation Tools
Confirmation Policy
Options
Five Dimensions
Galleries
File Structure
Workflow
Progress Checklist
Flow
Step 0: Load Preferences ⛔ BLOCKING
Step 1: Analyze Content
Step 2: Confirm Options ⚠️
Step 3: Create Prompt
Step 4: Generate Image
Step 5: Completion Report

Key facts

Author: jimliu (Baoyu)
Repo: jimliu/baoyu-skills
Installs on skills.sh: 32,289 as of Nov 2, 2026
GitHub stars: 27,159 for the whole jimliu/baoyu-skills repo, Nov 2, 2026
License: MIT
Last repo commit: Nov 1, 2026
What is inside: 35 files including SKILL.md: palette, rendering, dimension and workflow guides
skills.sh audits: Gen Agent Trust Hub warn, Socket pass, Snyk pass
Makes: A cover image for an article, 16:9 by default (also 2.35:1, 4:3, 3:2, 1:1, 3:4)
Choices: 6 types, 11 palettes, 7 renderings, 4 text levels, 3 moods, 4 fonts
Needs: An image backend: baoyu-image-gen with a provider key, Codex's imagegen, or Cursor's GenerateImage
Output folder: cover-image/<topic-slug>/ by default, with prompts/cover.md and cover.png

What it includes

Reads EXTEND.md; if missing, runs a first-time setup before anything else
Setup asks preference questions and saves EXTEND.md.
Try: “Make a cover image for article.md.”
Write · Load preferences (blocking)
Reads your text, saves reference images, picks an output folder
Looks for topic, tone, keywords and visual metaphors.
Analyze content
A hard gate: you confirm type, palette, rendering, text, mood and font
Skipped only with --quick, quick_mode or explicit wording; aspect ratio is still asked.
Confirm options
Saves prompts/cover.md from a template before any image call
Create the prompt
Sends the prompt to the chosen backend and saves cover.png
Retries once on failure. Existing cover.png is backed up first.
Runs code · Generate the image
Flags for each choice, plus aspect, language, quick and reference images
--type, --palette, --rendering, --style, --text, --mood, --font, --aspect, --lang, --quick, --ref. --no-title is the same as --text none.
Options
hero, conceptual, typography, metaphor, scene, minimal
Types
11 palettes (warm, dark, macaron) and 7 renderings (flat-vector, painterly, chalk)
Palettes and renderings
--ref files copied to refs/; people in refs are passed directly or described in the prompt
Files · Reference images
Regenerate or change one dimension: back up, update the prompt first, regenerate
Image modification
40 to 60 percent whitespace, simplified silhouettes, never invent the title
No realistic humans; the title is the exact one from you or the source.
Composition principles

From the official docs

Permissions

Medium risk

Claude sends your article text to an image provider and spends its credits, and it writes files into your project. Confirm the options and keep private material out.

What it can do

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

Watch out for

Image generation uses your provider credits
Each image is a paid call on the provider whose key you set. Check the confirmed options before approving, since regenerating text fixes means new calls. github.com
Your content is sent to the image provider
The prompt includes your source text and any reference images you pass. Do not feed it private material you cannot share with that provider. github.com
It writes files and backs up an old cover
It saves outputs in a new folder and, when you regenerate, backs up the existing cover.png first. Check the output folder before you commit. github.com
It runs backend scripts through bun
baoyu-image-gen runs TypeScript scripts with bun, or fetches bun with npx -y bun. Read the first command Claude proposes. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	SKILL.md defines no read-only mode. It confirms type, palette, rendering, text, mood and font with you before generating, unless you explicitly skip confirmation.
Talks to	The image provider you pick, through baoyu-image-gen: your prompt and any reference images. Package downloads if bun is fetched with npx -y bun
Sign-in and keys	Provider API keys live in .env files you create: ~/.baoyu-skills/.env or .baoyu-skills/.env in the project. The README says not to commit them.
Scopes	Claude saves source copies, analysis, prompt files and the image in your project Claude may copy reference images you give it into a refs folder Claude runs the image backend's scripts with bun, or npx -y bun if bun is missing A first run asks preference questions and saves an EXTEND.md file

Questions

What does baoyu-cover-image do?

It makes Claude generate a cover image for an article. Claude reads your text, proposes a type, palette, rendering, text level and mood, confirms with you, then generates the image through an image backend.

How do I install baoyu-cover-image?

Use the install box at the top of this page. In Claude Code it installs baoyu-image-gen with it, because the skill needs an image backend to make the picture. You also need one image provider API key. In Codex the skill uses Codex's own image tool.

Does baoyu-cover-image need an API key?

In Claude Code, yes. SKILL.md names native image tools only for Codex (imagegen) and Cursor (GenerateImage), so in Claude Code the skill uses baoyu-image-gen, which calls an image provider with a key you put in a .env file. It can also use a logged-in Codex CLI. In Codex or Cursor the skill uses the app's own image tool.

Can it add my article title to the cover?

Yes. The default text level is title-only, and the skill uses the exact title from you or the source and never invents one. Use --no-title for a pure visual.

Does it ask before generating?

Yes. By default it confirms the options with you first. It skips that only when you say so explicitly, for example with a no-confirm or quick option.

What if the text in the image is wrong?

It regenerates from a corrected prompt in a new file and keeps the flawed one for comparison. The skill forbids painting over or editing the text in a generated image with code.

Does it work in the Claude app?

No documented route. It needs scripts and an image backend on your machine, so the install box has no Claude app tab. Use Claude Code or Codex.

Is baoyu-cover-image safe?

It sends your article text and any reference images to an image provider you choose, spends that provider's credits and saves files in your project. Confirm the options before it runs and keep private material out.

Information

updated Nov 2, 2026

Repository: jimliu/baoyu-skills
Folder: skills/baoyu-cover-image

jimliu/baoyu-skills skills.sh

Sources

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

Show all 5

baoyu-skills README · github.com
skills CLI commands · github.com
baoyu-image-gen SKILL.md · github.com
baoyu-cover-image SKILL.md · github.com
skills.sh listing · skills.sh