> ## Content Index
> Fetch the complete content index at: https://claudefox.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# ai-manju-skills: 21 AI comic-drama templates that turn one idea into a full series plan
- URL: https://claudefox.com/ai-manju-skills-21-ai-comic-drama-templates-that-turn-one/
- Published: 2026-11-17T19:20:01.000Z
- Updated: 2026-11-17T19:20:10.000Z
- Description: ai-manju-skills is a set of 21 Chinese-language AI comic-drama templates that turn a single story idea into a production plan, asset prompts, episode scripts, storyboard and video prompts, and a release pack.
- Author: hupicake
- Tags: Tools

zhanghao123vip/ai-manju-skills gives a reader who wants to make an AI comic drama, known in Chinese as a manju, a complete starting document instead of a blank prompt. The reader supplies one story idea, and the chosen template works through a proposal, an asset library, episode scripts, a director handbook for each episode, and a release pack. The project contains 21 templates, one for each genre, and each template is a single SKILL.md file. There is nothing to install, because the templates are text files that an AI assistant reads and follows.

## What it is

Each template covers the full workflow for one series. The output starts with a proposal that sets out the story, characters, world and rules, episode-by-episode plot, foreshadowing, art, sound and release plans. It then builds an asset library of image prompts for characters, props, scenes and effects, followed by episode scripts with scenes and word-for-word dialogue. For each episode, a director handbook adds that episode's asset images, storyboard prompts and time-segmented video prompts. The final output is a release pack with cover prompts and publishing copy. The default shape in the README is 30 episodes, each about 4 minutes long, in 9:16 vertical format, and the reader can specify something else in the story idea. Shot counts and lengths are not fixed: each genre has its own rhythm, and each episode is adjusted to its content.

The 21 genres cover fantasy, historical and modern stories, and genre pieces. Fantasy includes xuanhuan cultivation, western fantasy magic, interstellar mecha, system and infinite-flow stories, and apocalyptic wasteland stories. Historical genres include time travel and rebirth, palace intrigue, household intrigue, court strategy, martial-arts jianghu, and farming and business stories. Modern genres include urban comeback, CEO romance, family drama, rebirth and revenge, workplace and business war, period drama, entertainment industry, and campus youth stories. The remaining two are suspense and crime investigation, and folk horror tales.

## Who it is for

The project is for a reader who wants to plan and produce a vertical short drama with an AI assistant, and who is comfortable working in Chinese, because the README and the templates are written in Chinese. It fits three setups that the README names. In Claude Code, the reader copies a genre folder into the skills folder and sends a story idea. In WorkBuddy, the reader imports the folder when that version supports importing a skill, and otherwise pastes the whole template text as in a chat tool. In DeepSeek and other chat tools, the reader copies the whole SKILL.md file, pastes it into a new conversation, waits for the ready reply, and then sends one story idea. One series uses one conversation, and one conversation uses one template, because mixing templates can mix up the settings.

## Quick facts

- Format: 21 SKILL.md template files, with no dependencies to install and no build configuration in the repository.
- Working language: Chinese, in the README and in the templates.
- Licence: MIT, with copyright in 2026 to zhanghao123vip, and use granted free of charge under the licence terms.
- Popularity: about 900 GitHub stars as of November 17, 2026.
- Releases: there are no releases as of November 17, 2026, so there is no release version to install.
- Defaults: 30 episodes, each about 4 minutes long, in 9:16 vertical format, unless the reader specifies otherwise.

## What stands out

- A five-stage workflow in one template: a proposal, a full-season asset library, episode scripts, a director handbook for each episode, and a release pack for each episode, so the reader does not have to re-explain characters, settings or rules at each step.
- Batched output with progress labels: the content is long, so it is written in batches, each batch starts by stating its progress, and the reader writes Continue to get the next batch, moving to the next stage when the current stage is finished.
- An asset library built for consistency: characters, props, scenes and effects each get copy-ready prompts, and characters get three to four unchanging anchor features, such as a scar position or a hair accessory, that later episodes must keep.
- A director handbook that ties images to video: each episode lists its asset images, gives storyboard prompts of about 100 to 200 words in the xuanhuan template, and gives video prompts split by time, with reference images anchored by name and number so the same character, scene and prop stay recognisable.
- A release pack for each episode: in the xuanhuan template it includes cover prompts in 3:4 vertical and 4:3 horizontal formats, a three-second ending card and an eight-to-ten-second preview, three hook titles, an 80-to-120-word introduction, five to eight hashtags, and a pinned comment.

## Quick start

From the project's README, in DeepSeek and other chat tools, open the SKILL.md file for the genre you want, copy all of its contents, open a new conversation, paste it and send it, wait until it replies that the template is ready, and then send one story idea. In Claude Code, copy the genre folder you need to \~/.claude/skills/ and then send your story idea directly. In WorkBuddy, import the folder if your version supports importing a skill, and if it does not, paste the whole template text as in a chat tool. Use one conversation and one template for each series. The instructions are on the README page at https://github.com/zhanghao123vip/ai-manju-skills, and there is no release to install from.

## How it was built with Claude

The repository shows Claude in its commit history: the commits carry a Co-Authored-By trailer naming Claude Opus 5.5\. The file tree contains no CLAUDE.md file, no AGENTS.md file and no .claude folder. The latest commit is dated November 1, 2026, and its message describes giving each manju skill its own pacing and filling workflow gaps.

**Try it with your AI.** Paste this into your AI assistant after you have copied the template text for the genre you want, using only the setup the README describes.

```
I want to try the ai-manju-skills template for [genre, for example xuanhuan-manju]. I
have copied the full SKILL.md for that genre from the repository
zhanghao123vip/ai-manju-skills. I will paste it into a new conversation in [DeepSeek,
Claude Code, or WorkBuddy]. If I am using Claude Code, I have copied the genre folder to
~/.claude/skills/. If I am using WorkBuddy and my version supports importing a skill, I
will import the folder; if it does not, I will paste the whole template text. Use only
the official commands and interfaces named in the README. Show risky steps first and
wait for me to confirm before you continue. Do not install anything from unknown
sources. Do not touch credential files. Do not delete anything. My story idea is: [one
sentence]. My defaults are [30 episodes, each about 4 minutes, 9:16 vertical, or your
own episode count, length and format]. Start with the proposal stage only, show the
progress label for each batch, and stop after the first batch so I can check it before
you continue.
```

## Known limits

- Testing is uneven: the README says the current version has been tested in WorkBuddy, the xuanhuan cultivation version has been through actual use, palace intrigue, household intrigue, period drama, system infinite-flow, CEO romance, rebirth and revenge, and workplace and business war have had an installation check or a stage-one trial run, and the remaining genres have not yet been tested one by one.
- The templates write prompts and scripts; they do not themselves draw pictures or render video, so the reader still needs separate image and video tools for the asset, storyboard and video prompts.
- Video-tool reference writing varies: the prompts use {{Mixed N}} as the reference-image writing, and the README says to change it to the corresponding writing when switching tools.
- Before publishing, the reader must check the rules of the platform where the work will be published and label the work as AI-generated content, and suspense and crime investigation and folk horror tales have more platform requirements, so the reader should confirm those first.
- The templates require characters and settings to be wholly original, and the README says the copyright and compliance responsibility for generated content rests with the user.
- In the xuanhuan template, blood is kept restrained, with no dismemberment, no disembowelment and no close-ups of wounds, and plots about taking a spiritual root or possessing a body are shown with spiritual-energy imagery rather than organs.

## Related on claudefox

- [Researcher says released weights show six Chinese models were started on Nvidia Blackwell GPUs](https://claudefox.com/checkpoint-forensics-chinese-models-blackwell-fingerprint/)
- [DeepSeek opens its app's search, file, voice and safety features to the API](https://claudefox.com/deepseek-app-api-search-ocr-asr-tts-pricing/)
- [rea-video: an open-source CLI that reverse-engineers a video shot by shot](https://claudefox.com/rea-video-reverse-engineer-video-shot-by-shot/)
- [huashu-motion lets Claude Code or Codex make a product launch film with music that you edit by clicking a frame](https://claudefox.com/huashu-motion-agent-launch-film-skill/)

## Problems and feedback

Report problems and ask questions on the Issues page at https://github.com/zhanghao123vip/ai-manju-skills/issues, where the maintainer can see them alongside the repository.

Source: [ai-manju-skills repository](https://github.com/zhanghao123vip/ai-manju-skills?ref=claudefox.com), [README](https://github.com/zhanghao123vip/ai-manju-skills/blob/main/README.md?ref=claudefox.com), [Releases](https://github.com/zhanghao123vip/ai-manju-skills/releases?ref=claudefox.com), [LICENSE](https://github.com/zhanghao123vip/ai-manju-skills/blob/main/LICENSE?ref=claudefox.com), [xuanhuan-manju SKILL.md](https://github.com/zhanghao123vip/ai-manju-skills/blob/main/skills/xuanhuan-manju/SKILL.md?ref=claudefox.com), [Issues](https://github.com/zhanghao123vip/ai-manju-skills/issues?ref=claudefox.com)

This story first appeared in the [claudefox daily briefing for November 18, 2026](https://claudefox.com/daily-2026-11-18/).