OnlyShot skill
A Claude Code skill that turns one sentence into a vertical AI micro-drama, from script and storyboard to video
- 341 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The scripts call the Dreamina CLI and spend paid Jimeng credits on frames and video
- The skill runs Python console programs and writes many files into the project folder
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/onlyshotDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The scripts call the Dreamina CLI and spend paid Jimeng credits on frames and video
- The skill runs Python console programs and writes many files into the project folder
This is third-party code. Review the repository files before installing.
What it does
OnlyShot is a skill that walks Claude Code through a production pipeline for a vertical micro-drama. You write one sentence with the idea, and the skill assembles a 24-episode script, a storyboard, reference images and video clips that are left for you to edit. The project targets Chinese micro-drama platforms and the Seedance 2.0 video model in the Jimeng service. The work runs in three stages. In the creative stage the skill studies the market, writes a brief and a story structure and drafts bibles for characters, scenes and props. It also builds a beat map down to the second and generates reference images. In the second stage each shot gets one to four still frames. The chosen frame later becomes the start of the video, so fixing a still costs 18 times less than fixing a clip. In the third stage scripts send the frames to Seedance through the Dreamina CLI and assemble video segments in one of four modes, picked by which files exist. The repository holds SKILL.md with the step-by-step flow, templates, screenwriting references and Python scripts. Final editing, voice-over and music are yours to do, for example in CapCut.
Who it is for. For authors and studios of short vertical dramas who work with Jimeng and Seedance and prepare clips for platforms such as Douyin.
Good fit when
- You have a series idea and want the full path to clips: script, storyboard, reference frames and video in one project
- You want to test a theme against the market: the skill searches what gained views in the last 30 days and steers away from worn plots
- You need characters and scenes to stay identical across episodes through bibles and large reference sets
- You want cheap composition fixes: stills at 3 credits each change first, and video runs only after that
- You need ready Jimeng prompts that account for known rejection causes
Not a fit when
- You only need a script: the skill is built for the whole pipeline and assumes video generation in Jimeng
- You have no Jimeng account and no Dreamina CLI: without them neither video nor reference frames get created
- You need horizontal videos or Western formats: templates are built for vertical 9:16 and Chinese platforms
- You want finished editing and voice-over with no manual work: the skill leaves assembly, voice and music to you
Example request
I want to make a short drama: an overworked programmer is reborn as a dog, and in episode one a system against his boss's pressure kicks in.Limitations
Claude Code on Windows, macOS or Linux and the Dreamina CLI, the official Jimeng tool, are required. It installs with `curl -fsSL https://jimeng.jianying.com/cli | bash` and signs in through `dreamina login`. On Windows, run the commands in Git Bash; PowerShell does not work. Video and frames are paid in Jimeng credits: one still frame costs 3 credits, a video segment about 55.
The materials and rules are made for a Chinese audience, platforms and language. The documentation and many of the files are written in Chinese. The skill's advice relies on the author's 2026 data, so success on a platform remains uncertain. Jimeng may reject a prompt under its moderation rules.
How to disable. Delete the skill folder: `rm -rf ~/.claude/skills/ai-short-drama` for a global install or `.claude/skills/ai-short-drama` inside a project. Projects with frames and video stay in your folders and must be removed separately.
FAQ
How do I install OnlyShot in Claude Code?
Clone the repository into the skills folder: `git clone https://github.com/A-cat-with-carrots/OnlyShot-ai-short-drama-skill ~/.claude/skills/ai-short-drama`. For a single project, clone into `.claude/skills/ai-short-drama`. Then restart Claude Code or open a new session and the skill loads automatically.
What does OnlyShot need to generate video?
A Jimeng account and the Dreamina CLI. The scripts use it to generate reference images, frames and video, and the multimodal mode uses Seedance 2.0 Fast. The output format is vertical 9:16 at 720p.
How is the storyboard stage different from plain prompt generation?
The skill first makes still frames from an eight-block description, you pick the best one and it becomes the first frame of the video. Composition is locked before the expensive generation. Key moments get a first-and-last frame pair.
How many episodes and shots does one project produce?
The skill is built around a 24-episode script. Each episode has 36 shots of 4 to 10 seconds with varied length, and the pacing is tied to seven retention beats placed between zero and 180 seconds.
How does the skill avoid Jimeng rejections?
Before generation the scripts check the prompt against the 1500-character limit, scan for words that trigger rejection and detect missing reference images. The reference lists 17 failure types and replacements for problem words.
Discussion
No questions or reviews yet.
Related
SkillsThree Claude Code skills: a 25-episode script, characters and storyboards, frame generation and video job submission
A set of 11 skills for Claude Code and Codex: from an idea or a novel to episode scripts, storyboards, image and video prompts, and a finished cut
A skill that teaches an agent to write prompts for ByteDance's Jimeng Seedance 2.0 video model: @ references, camera language, timing, templates
A skill that makes your agent film a short product video: concept, edit, sound and render on Remotion or HyperFrames
A skill for Claude Code and Codex that builds cinematic product videos with Remotion from shot recipe cards, a ready template and audio assets
Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| OnlyShot (this entry) | no | 341 | |
| Drama Skills | no | 3,479 | a skill set for screenplay and episode development, without its own video generation pipeline |
| ai-shortfilm-prompts | no | 499 | builds a five-stage prompt for video models, while OnlyShot drives the whole project through to video segments |
README in short
The documentation details the table of four video modes, the seven retention beats and the project folder layout with episodes, frames and clips. It also covers how the skill plugs into Jimeng and Seedance 2.0 and which checks run before generation.
README badge
Original SKILL.md
--- name: ai-short-drama description: 精细化 AI 短剧 IP 创作技能（v0.6.0）。三阶段架构：Phase 1 创作（剧本+ref图，反复迭代）→ Phase 1.5 分镜图（每 grid 1-4 张候选静态图 = 视频首帧，工业级核心层）→ Phase 2 出片（按集解锁，4 模自动选）。v0.6.0 关键升级：① ref 库工艺偏置铁律（现代摩天楼易拟物，古建筑必出 chibi 人体，前期 IP 设计阶段就要避坑）② 即梦 5.0 失败模式 + 敏感词清单（3 类 fail 区分 / prompt 1500 字硬上限 / 反派词替换表 / 暧昧词清单）③ 分镜图 8 段 prompt 模板（CHARACTER/BACKGROUND/ACTION/SCENE/CAMERA/LIGHT/TEXT/STYLE）+ NOT humans 子句必加。沿用 v0.3.0 升级：Phase 1.5 分镜图层、4 模视频、ref 5-8 最优。v0.2.0：bash → Python subprocess、36 grid × 4-10s 变奏、红果必爆 7 招、工业级 ref 库 80-150 张、单 prompt 300-500 字。务必触发：用户提到短剧、微短剧、竖屏剧、AI 短剧、AI 漫剧、剧本创作、分镜、即梦/Seedance 出片、红果/番茄/抖音 IP 改编、爽剧、重生、穿越、赘婿、追妻、神医相师、AI 漫剧奇观、或"帮我做一部短剧"类请求。 --- # AI 短剧精细化 IP 创作 v0.6.0 ## v0.6.0 changelog（2026-05-16） > SD-002《城市恋综》EP01 36 grid × 3 轮（v2.5/v2.6/v2.7）跑 + 全审反推。从「16/36 灾难级一致性破损」修到「0 灾难 / 60% ≥4 分」+ 沉淀 2 个工业级 SOP。 ### 关键升级 - ⭐ **`references/storyboard-frame-industrial.md`** v0.3.0：分镜图工艺 SOP - 8 段 prompt 模板（CHARACTER/BACKGROUND/ACTION/SCENE/CAMERA/LIGHT/TEXT/STYLE） - **NOT humans 子句必加**：每个主角单独 `Xxx body IS [arch] architecture NOT a human wearing X NOT a human with headpiece` - **ref 库工艺偏置铁律**：现代摩天楼易拟物 / 古建筑必出 chibi 人体 - 多角色 SOP：1 主角 ref + N 文字 silhouette + bokeh blur - 戏剧光词库（25+ 关键词）+ 红果级首帧 10 条 checklist - 11 节完整文档（含 EP01 v2.7 实测验证 grid13/21/22 全 5/5） - ⭐ **`references/jimeng-failure-modes.md`** v0.1.0（新）：即梦 5.0 失败模式 + 敏感词全清单 - **3 类 fail 完全区分**：InvalidNode（字数超） / generation failed（审核） / 无 image_url（网络） - **prompt 字数硬上限：1500 字符**（实测分布表） - **反派词替换表**：sinister/menacingly/dark intent → moody/leaning/quiet - **暧昧词清单**：blush/cheeks/extreme close-up → rose tint/medium shot - **并发 vs 串行限流**：长 prompt + 并发 = fail，长 prompt 必单跑 - 4 个 grid 实战诊断 case + 工艺校验 checklist - ⭐ **scripts/生成分镜图.py** 加 `--no-fingerprint` flag - 防 02_IP简报.md 的 ref 风格视觉指纹（neutral / flat / NOT cinematic）污染分镜图 dramatic 光 - max_workers 8 → 4（降并发防限流） - timeout 300 → 420s + 重试 2 次（防长 prompt 超时） ### 最大教训 ``` ref 工艺是分镜图工艺的上游。 ref 不纯（chibi 人体），分镜图加 NOT humans 也救不回来。 前期 IP 设计阶段就要避坑：选玻璃塔 / 简单几何，避复杂古建筑。 ``` --- ## v0.3.0 changelog（2026-05-08）