story-cover Skill: Design and Generate a Web Novel Cover With Your Agent

Turns a book title, pen name and genre into a cover prompt, and can then generate the cover image, sized for the platform you publish on.

Installs: 17K; on skills.sh
Stars: 8.2K; on GitHub
Files: 2; besides SKILL.md
Updated: Nov 2026; last repo commit

You give it a title, pen name and platform, and it picks a genre style, builds a prompt (text layer, style layer, scene layer) and offers two or three compositions. Say you only want the prompt and it stops after the prompt step.

To make the image it uses Codex's built-in ImageGen when available. Otherwise it falls back to the OpenAI image API and needs a GPT_IMAGE_API_KEY, plus curl, jq and base64. The skill text is written in Chinese and expects the Chinese novel folder layout that its sibling skills create.

Good for

Chinese web novel authors who draft with a coding agent
Writers working on Qidian, Fanqie, Jinjiang, Qimao or Zhihu Yanxuan
Authors who need a cover prompt to hand to a designer or another image tool

Install story-cover

Claude Code

Copy this prompt.

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

npx skills add zenstory-ai/oh-story-claudecode -y -g

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.

This installs the whole oh-story set of 14 skills, as the README says. story-cover reads the book folder that the other skills create, so run /story-setup in your writing folder and start a new session. Drop -g to install into the current folder only.

From the project README. Source.

What's in the skill

When it loads

小说封面生成与提示词设计。根据书名、作者名和题材构建封面提示词；用户只要提示词时不生图，要求封面图时调用 GPT-Image-2，Codex CLI 优先使用内置 ImageGen。触发方式：/story-cover、/封面、「帮我做个封面」「生成封面图」「封面提示词」「封面设计」。

Sections in SKILL.md

生成通路
生成流程
Step 1：收集信息
Step 2：题材判定
Step 3：构建提示词
Step 4：选方案并生成
Step 5：导出平台上传尺寸（平台有固定像素时）
Step 6：质量检查 + 迭代
参考资料

Key facts

Author: zenstory-ai
License: MIT (README badge)
Installs on skills.sh: 17,026 (Nov 2, 2026)
GitHub stars: 8,223 (Nov 2, 2026), whole repo
Files: 3 files including SKILL.md
Last commit: Nov 1, 2026 (repo)
Needs for images: Codex ImageGen, or GPT_IMAGE_API_KEY with curl, jq and base64
Language: SKILL.md and references are in Chinese
Upload size: Fanqie 3:4, 600x800; other platforms 2:3

What it includes

Reads the book folder for title, genre and platform; asks for a missing pen name
Looks for the current book (.active-book or a folder with tracking, text or settings), reads title, genre and target platform, and asks only for what is missing. Title and pen name are required and never invented.
Files · Step 1: collect info
Chooses a style from references/cover-styles.md
Matches the book's genre to the style library, or infers it from the title and tells you.
Read · Step 2: pick a genre style
Text, style and scene layers in one English prompt
Keeps the title and pen name as you wrote them, adds platform style keywords, and aims for a safe centre area. Prompt-only mode ends here with 1 to 3 prompts.
Step 3: build the prompt
Picks 2 or 3 compositions, then makes one image
Uses Codex built-in ImageGen when available and saves cover_vN files with a prompt text file. Falls back to the image API only when there is no built-in tool or you ask for it.
Write · Step 4: generate
Crops and resizes to the platform size
Runs a shell snippet with ImageMagick, or sips on macOS, to centre-crop to the exact upload size such as 600x800 for Fanqie. The original is kept.
Runs code · Step 5: export upload size
Checks text, genre fit, composition, platform and size
A checklist; if it fails, the agent regenerates with a new composition, colour or font.
Step 6: quality check

From the official docs

Permissions

Medium risk

What it can do

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

Watch out for

The API fallback can spend credits
If you pick the API route, each image request goes to OpenAI or your proxy with your key, and a reference image is uploaded to its edits endpoint when you give one. Use a key with a spending limit. github.com
It runs shell commands on image files
The export step runs ImageMagick or sips on the generated cover; the fallback uses curl, jq and base64. Check the output folder if the result looks wrong. github.com

Technical details

Talks to	OpenAI image API, or the proxy in GPT_IMAGE_BASE_URL (fallback only)
Scopes	GPT_IMAGE_API_KEY (API fallback only)

Questions

What does story-cover do?

It designs a web novel cover prompt from your title, pen name, genre and target platform, and can generate the image. If you only want the prompt, it stops after the prompt step.

How do I install it?

Use the install box above. The skill works as a single folder. The README installs the whole oh-story set of 14 skills with one command, which also adds the skills that create the book folder this one reads.

Does it need an API key?

Only for the API fallback. In a Codex CLI session with built-in ImageGen it needs no GPT_IMAGE_API_KEY. Otherwise it needs that key, and the fallback tells you not to run curl in the Codex path. The skill says it will not silently switch to a paid API if the built-in tool errors.

Can it make a cover without my book files?

Yes. With no book folder it saves covers under ./covers/<title>/封面/ and asks for the title, pen name and platform. In prompt-only mode it asks you for the genre if it has nothing to go on; otherwise it infers one from the title and tells you.

Is it safe?

It writes image files and a .prompt.txt file into your book folder (or ./covers). In API fallback mode it sends your prompt, and a reference image if you give one, to the OpenAI image API or the proxy in GPT_IMAGE_BASE_URL using your key, which may cost money on your account. The size step runs a shell snippet on the generated image. Check where the agent saves files and which key it uses.

Does it work in the Claude app?

The README documents agents that read project files, such as Claude Code and Codex CLI. Image generation needs either Codex ImageGen or the API key with curl, jq and base64, per the skill.

Do I have to read Chinese?

No. SKILL.md and its reference files are in Chinese, and the delivery message template inside it is Chinese too. It sets no reply language, so tell the agent to answer in English. The title and pen name are kept exactly as you give them, and the image prompt itself is written in English.

Information

updated Nov 2, 2026

Repository: zenstory-ai/oh-story-claudecode
Folder: skills/story-cover

zenstory-ai/oh-story-claudecode skills.sh

Sources

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

Show all 3

oh-story-claudecode README · github.com
story-cover SKILL.md · github.com
story-cover api-fallback.md · github.com