video-production-skills: AI Skills for Motion Design and Video Production Agents

What video-production-skills Is and Its Long-Term Goal

video-production-skills is maintained as a long-term skills library for AI video production workflows. The repository's README states its purpose is to collect reusable methods for video creation, reference recreation, motion design, opener packaging, and delivery quality assurance, and to decompose motion patterns from reference videos into HyperFrames and Remotion components that can be parameterized and reused.

Each top-level directory in the repository is a separate, independently installable skill. The README describes the long-term goal as building a video motion component library: each time a reference video is recreated, the resulting aligned motion patterns should be documented as technical components with their use case, input parameters, timing contract, implementation stack, alignment evidence, and limitations. Future AI video generation could then select from these components rather than generating static, slide-like sequences.

The last push to the repository was on July 14, 2026. The repository has no GitHub releases. No license file is present in the repository.

The Four Current Skills and Their Production Stage

The README's table maps each skill to a production stage and video style. The four current skills are:

ai-motion-director handles topic-to-motion direction and anti-PPT quality assurance. Given a theme, script, or brief, it establishes a motion thesis, beat graph, component and asset plan, and then routes tasks to HyperFrames, Remotion, or specific video skills.

reference-video-replica-qc manages the recreation loop. It decomposes a reference video's camera rhythm, entry style, path motion, component state changes, title layers, model lists, line tracks, card flips, light effects, masks, and transitions, then recreates them as parameterized HyperFrames or Remotion components with five-level fidelity classification.

dark-saas-magic-video is a styled production skill for dark SaaS and AI product shorts. The README describes the visual style as Presenton-like magic UI: black space, bottom purple glow, kinetic type, gradient CTA, floating UI elements, model rings, and export objects.

black-white-text-opener generates opening title cards with a pure black or near-black background, large white text typed out character by character, synchronized typing click sound effects, and a clean transition.

Installing and Choosing a Skill

Each skill is installed individually using the skills.sh CLI. To see all available skills in the repository:

npx skills add https://github.com/Pluviobyte/video-production-skills --list

Install a specific skill:

npx skills add https://github.com/Pluviobyte/video-production-skills --skill ai-motion-director
npx skills add https://github.com/Pluviobyte/video-production-skills --skill reference-video-replica-qc
npx skills add https://github.com/Pluviobyte/video-production-skills --skill dark-saas-magic-video
npx skills add https://github.com/Pluviobyte/video-production-skills --skill black-white-text-opener

The README provides a decision guide: use ai-motion-director when you want AI to establish the motion metaphor and beat graph before production begins. Use reference-video-replica-qc when decomposing an existing video and verifying alignment. Use dark-saas-magic-video for a dark tech product short. Use black-white-text-opener when adding a typed title card to a new video.

The repository structure follows the skills.sh convention: each skill directory contains a SKILL.md file and, for skills with supporting documents, a docs/ page. The README notes that docs pages for the four skills are at docs/ai-motion-director.md, docs/reference-video-replica-qc.md, docs/dark-saas-magic-video.md, and docs/black-white-text-opener.md.

reference-video-replica-qc: Five-Level Fidelity and Three Gates

The reference-video-replica-qc skill introduces a structured fidelity classification system with five levels. Level one is source code consistency (the code produces the same rendering instructions). Level two is rendered frame pixel consistency (individual frames match at the pixel level). Level three is encoded video frame alignment (the compressed output frames align). Level four is visual alignment (the motion and layout are visually equivalent). Level five is style alignment (the overall aesthetic matches).

The skill also defines three verification gates that any pixel-level claim must pass: the asset gate (source materials are present and unmodified), the runtime gate (the rendering environment produces the expected output), and the delivery gate (the final encoded file matches). The README states that evidence for pixel-level conclusions must come from MAE (mean absolute error), temporal offset measurements, boundary frame checks, PSNR/SSIM scores, and hash comparisons.

The practical effect is that a recreated video is not described as "matching" the original without specifying which fidelity level was reached and what evidence supports it. Manually written HyperFrames or Remotion recreations typically target visual or style alignment rather than pixel-level alignment.

Limitations and the Missing License

The most significant issue is the absent license. No LICENSE file exists in the repository, and the README does not include a license statement. Without a license, the default under copyright law is that all rights are reserved by the author. Using the skills in a commercial production pipeline or distributing derivative works based on the SKILL.md content may require explicit permission from the maintainer.

The repository currently has four skills. The README acknowledges that the collection is not finished: it describes six planned categories (direction and scheduling, recreation and component capture, openers and packaging, styled main videos, asset processing, platform adaptation) with additional skills to be added. The current four skills do not cover platform-specific requirements for TikTok, Xiaohongshu, or WeChat Video covers and ratios, which the README lists as a planned category.

The skills are written for HyperFrames and Remotion. If your video production pipeline uses a different programmatic video framework, the skills' output code may need adaptation.

video-production-skills vs. Unguided Agent Prompting

The simplest alternative to video-production-skills is writing your own CLAUDE.md or system prompt describing what kind of videos you want to produce and which framework you use. This requires no external installation and no license concerns, since the guidance is entirely your own.

video-production-skills offers pre-built guidance for specific video types. The reference-video-replica-qc skill provides a fidelity classification framework and evidence requirements that would take time to develop independently. The anti-PPT gate in ai-motion-director encodes an editorial standard (avoid static slide-like sequences) that most video prompts do not include explicitly.

The trade-off is maintenance and licensing. An in-house CLAUDE.md stays under your control, can be tuned for your specific tools and style, and has clear legal status. video-production-skills is a shared library that evolves as the author adds content, but requires trust that the SKILL.md files accurately encode the production methods described in the README.

Editorial conclusion

video-production-skills is a practical starting point for AI-assisted video production workflows that use HyperFrames or Remotion, particularly when an agent needs a structured motion thesis before generating code or when a reference video must be recreated with documented fidelity evidence. Its scope is specific: the four current skills cover dark SaaS product shorts, text openers, motion direction, and reference recreation. There is no license statement in the repository, which means the legal status of derivative works is unresolved. Check with the maintainer before using this repository in a commercial production pipeline.

Frequently asked questions

Official sources

1. Issues
2. Pluviobyte/video-production-skills on GitHub
3. README