Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Short-Form Video from HTML Scenes (1080x1920 MP4)

    1

    Screen-recording an animation drops frames on a slow machine and never comes out the same twice.

    Free

    Security scanned
    Short-Form Video from HTML Scenes (1080x1920 MP4)

    Short-Form Video from HTML Scenes (1080x1920 MP4)

    Example session with this skill installed

    숏폼 만들어줘: 카드뉴스를 HTML로 찍는 스킬 소개, 13초. (Make a 13-second vertical short that introduces the card news skill.)

    • Read your context and instructions
    • Compiled the short-form video from

    Summary

    Wrote a four-scene, 13-second script, built it as HTML scenes from the template, and rendered a 1080x1920 MP4 (390 frames at 30 fps) plus a cover JPG. The renderer reported no text outside the safe area.

    Scene table (shown before writing any HTML)

    SecondsOn screenJob
    0-3카드뉴스, 아직 디자인 툴로 만드세요?A question the viewer answers in their head
    3-6.4HTML 한 장이면 돼요 + a code cardThe answer, shown as code
    6.4-10.2이렇게 해요: 1 한 장에 한 메시지 · 2 카드마다 PNG로 찍기 · 3 넘침 · 잘림 자동 검사Three steps, one line each
    10.2-13파일은 깃허브에 + repository addressOne call to action

    You: 좋아

    Render

    node render.mjs short.html out/demo.mp4

    0%8%15%23%31%38%46%54%62%69%77%85%92%demo.mp4 · 13초 · 390프레임 · 21초 걸림
    표지 demo.jpg
    

    (13초 = 13 seconds, 390프레임 = 390 frames, 21초 걸림 = took 21 seconds on an M-series MacBook. 표지 = cover image.)

    Files: out/demo.mp4 (1080x1920, H.264, BT.709) and out/demo.jpg (cover taken at 1.2 s).

    Checked by eye

    • The cover (1.2 s) shows the opening question.
    • Frame 0 already shows the opening question: the first scene has no fade-in, and the renderer checks that frame 0 and the cover show text (no warning on this run).

    Next steps

    • Upload as a Short / Reel / TikTok with the cover set; put the link in the description or a pinned comment.
    • If you add an AI voice, turn on the platform's AI label.
    • For the next video, change the scene layout instead of only swapping the text.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Screen-recording an animation drops frames on a slow machine and never comes out the same twice. This skill renders a short video the way a game engine would: it moves a clock to each 1/30 second, takes a screenshot, and joins the frames with ffmpeg. The same HTML always makes the same video.

    What it does

    1. Script first. A scene table (seconds, on-screen text, voice line) is shown to you before any HTML is written. The first 0-3 seconds carry a question or a result; one idea per scene; one call to action at the end.
    2. Writes the HTML from the bundled template. A scene is <section class="scene" data-start="3" data-end="6.4">; motion is plain CSS animation.
    3. Renders node render.mjs short.html out.mp4 [--audio voice.mp3] into a 1080x1920 MP4 (BT.709) and a cover JPG taken at the time you choose.
    4. Warns about text outside the safe area (x 80-960, y 260-1540), where captions and platform buttons cover it, and when frame 0 (the feed preview) or the cover shows no text.
    5. Watches the result from start to end and checks that the first frame and the cover read on their own.

    Rules it follows

    • Does not mass-produce videos that only swap text in one template (platforms act on repetitive content). Each video gets its own scene layout.
    • No unverified numbers, superlatives or guarantees.
    • Turns on the AI label when an AI voice or realistic AI footage is used.
    • Music and footage only when you have the right to use them.

    Needs

    Node 18+, npm i playwright once (uses your installed Chrome when it can), and ffmpeg.

    Language

    Instructions and template copy are in Korean. Your agent can follow them and write scenes in any language.

    Source, example video and the other skills: https://github.com/Stophyuk/solutionzip-skills

    Instructions and template copy are in Korean. Rendering is frame by frame, not real time: the 13-second example took about 25 seconds on an M-series MacBook. It does not record a voice; bring your own voice.mp3 or TTS. Music and footage must be files you have the right to use.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 16 days ago

    • Free with an Agensi account

    Requires

    Node.js 18+
    playwright
    ffmpeg

    Permissions

    Browser
    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Listed16 days ago

    What's inside

    Frequently Asked Questions

    A screen recording drops frames when the machine is busy. Here the clock is moved to each 1/30 second before a screenshot, so a slow machine only takes longer; the video comes out the same every time.

    Yes. Pass an audio file with --audio voice.mp3 and line the scene times up with the voice. The skill does not make the audio itself.

    Node 18 or later, Playwright (npm i playwright), Chrome or Playwright's Chromium, and ffmpeg.

    Yes. Your agent reads the Korean instructions and can write the scenes in English or any other language. Rendering does not depend on the language.

    New in Workflow Automation