Agent skills that turn a shot idea, a script or a reference clip into a prompt a specific video model can follow: Seedance 2.0 and 2.5, Kling 3, Veo and Gemini Omni, MiniMax H3, Higgsfield. Many only write the prompt; some also run the generation.
81 skills
Start with the model you generate with, because prompt syntax does not carry over. Seedance has the most skills. Seedance 2.0 Skill OS is a routing skill with 28 sub-skills covering camera, audio, character locks, continuation and troubleshooting. Seedance 2.0 Prompt Writing Skills teaches @ reference roles and time-segmented prompts, in English and Chinese editions. Seedance 2.5 Video Director targets the newer model and only prepares prompts; it never submits a paid job. Compare each skill's last-updated date with the Seedance version you use: some Seedance 2.0 skills have not changed since early 2026.
For other models the field is smaller. Kling AI Prompt Generator helps choose a Kling model and writes prompts for image-to-video, text-to-video, multi-shot and talking-avatar work, while Kling 3.0 Prompting Skill asks for the generation mode before assembling a prompt. For Google's own models there is the official Gemini Omni Flash API skill, which pairs prompting notes on scenes, audio, timing and on-screen text with scripts that call Gemini Omni 1.1 Flash; for Veo prompts, use the cross-model skills below. MiniMax's official MiniMax H3 skills include one portable prompt-writing skill; its style-specific workflows need the MiniMax Hub agent. Higgsfield users can read Higgsfield AI Prompt Skill and Higgsfield Seedance 2.0 Skills; both write prompts you run on Higgsfield with your own account. There is no Sora-only skill in this list. Sora appears only as one target among several in the cross-model skills below.
If you switch between models, pick a cross-model skill. Video Prompting from Square-Zero-Labs asks which model and input mode you use, then follows that model's own guide (MiniMax H3, Seedance 2.0 and 2.5, Veo 3, Wan, LTX). Visual Skills: video plans each clip as drama first, then reads one model file for Seedance, Kling or Veo. lanshu-awesome-ai-video-kit adds a model picker and a translator that rewrites a prompt from one model's formula into another's; its instructions are in Chinese. For Sora, AI Shortfilm Prompts and Video Prompt Reverse both list it among their targets. To work backwards from a clip you like, Video Prompt Reverse splits the footage into shots, motion, lens, color and audio before writing, and MiniMax H3 Video Reverse states what it saw, waits for your confirmation and only then writes H3 prompts.
Character consistency is mostly handled in the prompt, so look at how a skill does it. Seedance 2.0 Skill OS has a character-lock sub-skill, and Seedance 2.5 Video Director records each reference's role in an asset lock table, including character and wardrobe locks. The character-design skill in fal.ai Skills separates a fixed character anchor from the variables that change per shot, and Replicate prompt-videos advises reusing one character description word for word. Film Studio Skills holds back prompting until every character, location and prop in the shot is locked. A written anchor is one part of keeping a face steady across shots; reference images are the other, covered in consistent character AI.
Before installing, check five things. Agent: some skills run in Claude Code, Codex and Cursor, others in only one; Kling AI Prompt Generator installs as a Claude Code plugin. Language: the Kling skills and the Gemini Omni Flash API skill here are in English, while many Seedance skills are Chinese or bilingual; each skill's page lists its content language and the agents it works with. License: MiniMax H3 skills, Seedance Prompt Skill and fal.ai Skills declare no license on GitHub, Visual Skills: video is CC BY 4.0 and requires attribution, and Huobao Drama Skills carries a non-commercial license (CC BY-NC-SA 4.0). Cost: many only write text, but fal.ai Skills runs through the genmedia CLI with a fal API key, Gemini Omni Flash API needs a Gemini API key, and Pexo Skills drives a paid hosted video agent. Freshness: model versions move quickly, so check the last push before trusting a prompt format.
Emily2040/seedance-2.0
A large skill set for writing Seedance 2.0 video prompts: a routing root skill plus 28 sub-skills for camera, audio, character locks, continuation and troubleshooting.
For Seedance 2.0 users who want one skill that routes between modes and covers camera, audio, character locks and troubleshooting. MIT-licensed, English and Chinese material, and it works in Claude Code, Codex, Cursor and Gemini CLI.
liyue-aigc/seedance-2-5-video-director
A Seedance 2.5 prompting skill that turns an idea, storyboard, reference media or an old prompt into a script, diagnosis or copy-ready prompt without submitting any paid generation job.
For Seedance 2.5 users who want a prompt checked before paying for a generation. It locks mode, duration and assets first, writes a timeline prompt and audits it without submitting any paid generation job.
maciejdzierzek/kling-ai-prompt-generator
Claude skill for writing Kling video prompts and choosing a Kling model, covering image-to-video, text-to-video, multi-shot storyboards, talking avatars and motion transfer.
For Kling users choosing between models and writing image-to-video, text-to-video, multi-shot or avatar prompts. It installs as a Claude Code plugin; its facts are dated 2026-08-04, so recheck prices and limits.
google-gemini/gemini-skills
Google's skill for generating, extending and editing video with Gemini Omni 1.1 Flash through the google-genai SDK, with scripts for upload, ffmpeg prep and batch runs.
For Google's own video model, from Google itself. It covers text-to-video, first and last frame transitions, extending a clip by 10 seconds per turn and editing a clip, plus prompting notes on single scenes, audio, timing and on-screen text. It needs a Gemini API key and Python, and uploading video for edits is unavailable in some regions.
MiniMax-AI/MiniMax-H3
MiniMax's official skills for its H3 video model: one portable prompt-writing skill plus eight style-specific video workflows that need the MiniMax Hub agent.
For MiniMax H3 users. The official h3-prompt-writing skill identifies the input mode and rewrites your request into H3's structure with no API calls; the other workflows need the MiniMax Hub agent.
Square-Zero-Labs/video-prompting-skill
A router-style skill that drafts video prompts in the format each model expects, plus a character-sheet workflow for locking a character's look before image-to-video.
For people who switch models or want character sheets first. It asks which model and input mode you use, follows that model's guide, and has a workflow for locking a character's look before image-to-video.
luozhilzh/video-prompt-reverse
Markdown skill that reverse-engineers a video or a rough idea into AI video prompts, treating camera and subject motion as a core dimension and adapting the result to 10+ video engines.
For recreating a look you already have a clip of. It splits footage by shot, motion, lens, color and audio, writes a reconstruction prompt with variants, and adapts it to 10+ video engines.
nexu-io/open-design
A skill bundled with the OpenDesign app that plans one short AI video shot, writes the prompt the way the chosen model prefers and renders it through the app's media command.
HBAI-Ltd/Toonflow-app
Skills that come with the open-source Toonflow canvas app: a production workflow with Seedance 2.0/2.5 prompt rules, a canvas manual, a text-storyboard skill and an MCP skill for outside agents.
chatfire-AI/huobao-drama
Seven skills from the Huobao Drama app, in Chinese with English variants: novel-to-script rewriting, asset extraction, storyboard segments, and character, scene, prop and video prompts.
waooAI/waoowaoo
Six Chinese-language skill files from the waoowaoo AI film workspace, covering script writing, creative direction, asset design, video-segment prompts and music cue planning.
MiniMax-AI/MiniMax-H3
MiniMax's official skills for its H3 video model: one portable prompt-writing skill plus eight style-specific video workflows that need the MiniMax Hub agent.
Emily2040/seedance-2.0
A large skill set for writing Seedance 2.0 video prompts: a routing root skill plus 28 sub-skills for camera, audio, character locks, continuation and troubleshooting.
google-gemini/gemini-skills
Google's skill for generating, extending and editing video with Gemini Omni 1.1 Flash through the google-genai SDK, with scripts for upload, ffmpeg prep and batch runs.
dexhunter/seedance2-skill
Teaches an AI agent to write prompts for Jimeng Seedance 2.0: @ reference roles, time-segmented structure, camera terms and ready-made templates, in English and Chinese editions.
eternityspring/shuohao-skills
Six skills for AI short drama that turn a novel into an adaptation outline, character sheets, an art bible, a script and a storyboard, with scripted quality gates at each stage.
songguoxs/seedance-prompt-skill
A single-file Claude Code skill that writes Chinese prompts for Seedance 2.0 on ByteDance's Jimeng platform, using its @ reference syntax and ten prompt patterns.
liangdabiao/Seedance2-Storyboard-Generator
Adapts a novel or story outline into a multi-episode AI short drama script, writes image prompts for numbered character, scene and prop assets, and drafts Seedance 2.0 storyboard prompts per episode.
MiniMax-AI/cli
Official skills for the MiniMax mmx command line: media generation, speech with subtitles, transcription, and a dedicated skill for running MiniMax-H3 video jobs.
LeoYeAI/openclaw-master-skills
Two skills picked from a large OpenClaw collection: a pipeline that turns a novel or idea into AI short-drama prompt packages, and a reference for calling the LTX-2.3 video API.
jd-opensource/JoyAI-Echo
Four skills from JD's JoyAI-Echo project: a director workflow, a story writer and a tail-frame prompt rewriter for multi-shot Echo 1.5 video, and a six-field prompt writer for Echo-WM.
nolanx-ai/nolanx.ai
Nineteen film-craft skills from the open-source NolanX runtime, covering story structure, Fountain screenplays, shot language, continuity and Seedance 2.0 prompts.
kaomei/stickman-video-director
Codex skill that turns copy, notes or an article into a confirmed director's proposal, then writes ten-second Gemini Omni Flash prompts for English stick-figure explainer videos.
lixiaoxiao9888-create/manju-laoli-skill
A Chinese-language director suite for AI short dramas that locks the video model and aspect ratio first, then runs gated scripting, asset locking, shot design and prompts with a built-in checker.
beshuaxian/higgsfield-seedance2-jineng
Fifteen Claude skills, each writing detailed Seedance 2.0 prompts for Higgsfield in one video style, from cinematic and anime to music video and product ads, with Chinese versions.
MapleShaw/seedance2.0-prompt-skill
A Chinese-language skill that writes video and image prompts for Seedance 2.0 on Jimeng, with a Z/Y/X/F camera code, a character-to-storyboard route for long videos and editing-rhythm formulas.
pexoai/pexo-skills
Open-source skills from Pexo for AI video: a client for its hosted video agent, a director skill, Seedance 2.0 and Veo prompt writers, and video and audio generation studios.
s1dashu/director
A top-level director skill that routes a video idea into one of three modes, then builds character and voice references, writes shot prompts and generates Seedance 2.0 clips.
liangdabiao/make-prompt-seedance2
Chinese-language skill that routes a short-video idea to one of five Seedance 2.0 prompt methods and returns a paste-ready, timeline-based storyboard prompt.
OSideMedia/higgsfield-ai-prompt-skill
A Claude skill library for Higgsfield video prompts: a dispatcher plus about 35 sub-skills on camera, Seedance, Cinema Studio, acting, story bibles, shotlists and troubleshooting.
agentara/skills
Four skills that take a short-video idea to a timed scene plan, a storyboard sheet with a matching prompt script, a character sheet and poster key art, with an image model drawing the pictures.
smixs/visual-skills
A video-prompt skill that has the agent plan each clip as drama (want, obstacle, staging, cuts) before writing the syntax for Seedance, Kling or Veo.
bytedance/agentkit-samples
Six skills from ByteDance's AgentKit samples repo: Seedance prompt writing and rewriting, Seedance video generation, a comic-drama pipeline, shot breakdown and FFmpeg editing.
freestylefly/canghe-skills
Four video skills from a larger personal collection: a Seedance manga-drama generator, a manga-style clip generator, a Seedance API runner and a Flyworks digital-human client.
Rimagination/h3lite
A Codex skill that sets up and runs MiniMax H3 locally through ComfyUI on a Windows NVIDIA PC, picks a component set that fits the GPU memory and generates short clips with native audio.
jnMetaCode/ai-shortfilm-prompts
Claude Code skill that turns an idea into a cinematic video prompt for models such as Seedance, Sora, Kling and Veo, using a five-part structure and 21 genre templates.
cclank/lanshu-awesome-ai-video-kit
Seven Claude Code skills for AI video prompts: a model picker, a cross-model prompt translator, Seedance, HappyHorse and Kling prompt writers, a Seedance storyboard splitter and a Seedance debugger.
liyue-aigc/seedance-2-5-video-director
A Seedance 2.5 prompting skill that turns an idea, storyboard, reference media or an old prompt into a script, diagnosis or copy-ready prompt without submitting any paid generation job.
rediumvex/ai-video-generator-claude
Ten Claude skills that each turn a one-line idea into a timestamped Seedance 2.0 prompt for Higgsfield, for formats such as viral hooks, faceless channels and SaaS launch clips.
mediastormDev/dream-to-video-skill
An agent skill that turns dream or diary text into a Jimeng video: it writes the prompt, queues it, drives the Jimeng web page with Playwright and downloads the result.
zhouwei713/seedance-prompt
Writes documentary-style video prompts that avoid the AI look by first fixing who filmed the footage, on what device and in what era, then filling a seven-part template.
elementsix/elementsix-skills
A Claude Code skill that walks you from a rough idea to a time-coded Seedance 2.0 storyboard prompt, with @-references to your image, video and audio files.
nutllwhy/seedance-tvc-director
Text-only director skill for 15 s, 30 s or custom-length commercial spots: turns a brand brief into concepts, a shot-by-shot sheet and copyable Seedance prompts, then plans retakes from the cut.
T8mars/minimax-h3-prompt-skill-T8
More than 280 installable skills that turn curated short-video cases into reusable creative mechanisms and write prompts for MiniMax H3 and Seedance 2.0, plus a case catalog and a desktop viewer.
liyue-aigc/xianxia-cinematic-video-director
A Codex skill that turns a xianxia (Eastern fantasy) scene idea, keyframe or reference clip into a multi-shot plan with camera moves, continuity locks and per-shot image-to-video prompts.
Hao0321/ai-media-generator
Picks a generation platform and writes platform-specific prompts for AI images, video and music, and can submit them by driving the platform's website through browser automation.
fal-ai-community/skills
Five fal.ai skills run through the genmedia CLI: cinematic prompting, multi-shot storytelling, consistent characters, model-specific prompt guides and multi-step pipelines.
LetMeHappyCode/auto-video-agent
Claude Code project that turns an SRT subtitle file into storyboard blocks, illustration prompts, Seedream images and a silent slideshow-style video.
gnipbao/minimax-h3-video-reverse-skill
Skill that turns a reference video into MiniMax H3 prompts: it restates what the video shows, waits for your confirmation, then writes text-to-video or per-shot image-to-video prompts.
calesthio/generative-media-skills
A large skill suite for AI media production: video-model operating guides, shot and storyboard direction, character continuity, editing and delivery checks across image, video and audio.
Square-Zero-Labs/video-prompting-skill
A router-style skill that drafts video prompts in the format each model expects, plus a character-sheet workflow for locking a character's look before image-to-video.
jijiutong/ai-visual-director
A Claude Code skill pack that turns a story into image prompts for character sheets, scenes and storyboards, plus video prompts for five platforms, with quality checks and locked-state tracking.
Supreme-Ultimate/novel-to-script-team
A multi-agent, multi-skill workspace that adapts novels into short-drama or serial scripts, then into storyboards and video prompts, with review gates between stages.
phileiny/h3-storyboard-skill
A Claude Code skill for splitting a script into MiniMax H3 shots and directing faces so expressions render, with each rule marked as verified, partly verified or inferred.
kangarooking/director-skills
Four director-style skills for AI video: fight-scene prompts, concept-art asset prompts, tourism promo planning that mixes real and generated shots, and a reliability layer for the Miora video tool.
wuwangzhang1216/DirectorSKILL
A director and previs skill that turns a script, prose or keyframe into a production plan, with an optional single-director style lens and a coded guide to repairing failed generations.
Alisa0808/vibe-creating-skill
A prompt skill that first judges whether a rough idea or shot script suits the Vibe Creating style, then rewrites it into a cleaner text-to-video prompt, asks for details, or leaves it as is.
wyuzhi/qisi-video-remix
Takes a reference video or story material and writes a new story with a full script, character and scene image prompts, a shot table, generation segments and copy-ready video prompts per segment.
machina-exm/film-studio-skills
Seven chained skills that take a script or idea to generation-ready shot prompts, with gates that hold back prompting until every character, location and prop in the shot is locked.
Rylaispirit/cinematic-video-prompt-skill
A cinematography vocabulary skill that helps an agent write specific text-to-video and image prompts, plus two small bonus skills for character reference sheets and xianxia-style scenes.
MasterLeos/leos-six-department-directing-team-skill-v1
A directing workflow for AI films and series in which six roles, from lead director to continuity supervisor, work from one shared scene brief before a video prompt is written.
zhanghaonan777/Seedance2-skill
Acts as a creative director for Seedance video: reads supplied images, expands copy into a prompt from a camera-term vocabulary, self-checks it and can call the Seedance API via a bundled script.
black-forest-labs/skills
Official Black Forest Labs suite of eight linked skills for FLUX 3 video: router, brief checker, prompt craft for shots, keyframes, audio and period looks, API runner and product-ad workflow.
woodfantasy/Seedance-ShotDesign-Skills
A skill that turns a video idea or an existing clip into a copy-ready Seedance 2.5 prompt, routing between generation, long-video, extension, editing and other Jimeng modes.
zyz254009-crypto/script-to-shootable-storyboard
Turns a script, synopsis or existing storyboard into one-action shootable shots with continuity and timing checks, then routes each shot to live action, Seedance or a hybrid.
SlavaSexton/ComfyUI-Agent-Kit
A local-first ComfyUI kit: a router skill that drives your own ComfyUI, plus standalone skills for Seedance, MiniMax H3 and Krea models.
heloraai/Seedance2.0-Prompt-Optimizer-skill
A skill with Chinese instructions that turns a one-line idea into a prompt for Jimeng's Seedance 2.0, using genre templates and a subject, camera, effect, light and audio formula.
zhouwei713/fpv-immersive-video-prompting
Prompting skill for first-person one-shot AI videos: it plans the camera route as numbered stops or a red-line path and writes the video prompt plus matching image asset prompts.
replicate/skills
Replicate's model-agnostic guide to prompting AI video models: scene description, camera language, audio and dialogue, time-coded multi-shot prompts, reference inputs and character consistency.
PixVerseAI/skills
PixVerse's own skill library that teaches an agent to drive the PixVerse CLI for video, image, speech and music generation, with Seedance prompt optimizers and reusable character references.
62656456/ai-film-skills
A set of independent skills for screenplay and director thinking, character, scene and prop assets, storyboards and video prompts, genre visual presets, and producing an actual video.
aicontentskills/ai-video-storyboard-skill
A SKILL.md that plans multi-shot AI videos longer than one generation as a shot list with a shared visual theme and a copy-ready prompt for each shot.
RainLib/AI-Storyboard
Four Claude Code skills for episode scripts, nine-panel and four-panel storyboard prompts, motion prompts for image-to-video tools and a pass-or-fail storyboard review.
senmanx/novel-to-manju
A self-contained ten-step skill that turns a novel or topic idea into a rewrite, a first-person manju (comic drama) script, character and scene prompts, a storyboard and MiniMax H3 video prompts.
luozhilzh/video-prompt-reverse
Markdown skill that reverse-engineers a video or a rough idea into AI video prompts, treating camera and subject motion as a core dimension and adapting the result to 10+ video engines.
LuoJiangYong/muse-video-skill
A pre-production skill that runs a virtual film crew to turn a video idea into a script, storyboard, art direction and model-ready prompts; it plans the video but does not render it.
alchaincyf/seedance-skill
A workbench for steering Seedance 2.5 with clay-model reference videos: clarify the shot, render a white model and write the prompt from one scene.json, with a geometry check before rendering.
zhouwei713/facial-expression-prompting
A skill that turns a vague emotion into a motivated, beat-by-beat facial and body performance and a complete AI video prompt, or a block to paste into an existing prompt.
maciejdzierzek/kling-ai-prompt-generator
Claude skill for writing Kling video prompts and choosing a Kling model, covering image-to-video, text-to-video, multi-shot storyboards, talking avatars and motion transfer.
mini-yifan/open-source-video-gen-skill
Nine skills that run an AI video crew on the open-source MiniMax H3 model and a rented AutoDL GPU, from H3 prompts and reference art to a gated short-drama pipeline, music and voice-over.
PostPlusAI/postplus-skills
Skills for the PostPlus CLI covering batch video generation, shot-by-shot reference analysis, reference decoding, transcription to SRT or ASS, and voice planning and execution.
anyhui/video-storyboard-gen-skill
A Chinese-language skill that turns a topic or idea into a story structure, a timecoded shot table and optional Jimeng and Seedance video prompts.
aedev-tools/kling-3-prompting-skill
Prompt-writing skill for Kling 3.0 that asks for the generation mode, collects scene details and assembles a director-style prompt.
krea-ai/skills
Krea's own skills for driving its image and video models through a connected Krea MCP server, with model picking, prompting guides, product motion, marketing assets and narrated explainers.
L-Trunks/ai-film-skills
Six directing skills for AI short films on local or cloud video models: opening decisions, character reference, shot screening, camera language, pacing and storyboarding from available footage.
whystrohm/shotkit
Five skills that turn a video brief, plus an optional brand-lock file, into a storyboard and shot list, prompts for ten image and video generators, an HTML preview and per-frame critiques.
Arch-Dog/video-prompt-engineer
Converts a finalized short-drama script fragment into Seedance-oriented 10 to 15 second text storyboards and an executable video prompt table.
fanchengchen1-collab/super-director
Thirteen director-craft skills for AI short drama: Chinese and English main systems for blocking diagrams, shot tables and prompts, a routing kit, nine director-style skills and a director library.
A SKILL.md instruction pack for a coding agent such as Claude Code or Codex. It teaches the agent how Seedance reads @ references, timestamps and camera terms, so you describe a shot and get a prompt in that format back. Many only write the prompt and leave generation to you; each skill's page says whether it also calls an API. Install steps are in the skills directory.
This page lists skills that write prompts, not libraries of finished ones. Some ship examples: Seedance 2.0 Skill OS includes example banks in Chinese, Japanese and Korean, and Seedance 2.0 Prompt Writing Skills has templates for an ad, a short drama scene, a dance clip and a scenery montage. For storyboard-style prompts with a timeline, see Seedance Video Prompt Skill and Seedance 2.0 Storyboard Generator. There is no Sora-only skill here; AI Shortfilm Prompts covers Sora alongside Seedance, Kling and Veo.
A few can. Video Prompting from Square-Zero-Labs loads a separate guide for each model, Visual Skills: video reads one model file at a time for Seedance, Kling or Veo, and lanshu-awesome-ai-video-kit adds a translator that converts a prompt from one model's formula to another's. A prompt tuned for one model should not be pasted into another unchanged.
Many do not. Seedance 2.5 Video Director never submits a paid job, and the Higgsfield Seedance 2.0 Skills write prompts you paste into Higgsfield yourself. Others do call a service: fal.ai Skills runs through the genmedia CLI with a fal API key, Gemini Omni Flash API needs a Gemini API key, and Pexo Skills drives a paid hosted video agent.
Write one fixed character description and reuse it word for word, keeping what changes per shot (pose, outfit, light) separate, as the fal.ai Skills character-design skill and Replicate's prompt-videos guide suggest. Reference images carry the rest of the work; Pencil0 uses them to keep characters consistent, as described in consistent character AI.
REEL END — 00:00:00:00
Pencil0 runs script, character and scene references, shots and the final cut for you, one confirmed stage at a time, and keeps characters consistent across shots with reference images. It is an invite-only beta.
Join waitlistFree to join · we'll email you the day we open