Skip to main content
skill CX Codex

scripted-screencast — Claude Code skill: demo recording

The scripted-screencast Claude Code skill turns scripted browser actions into a product demo video with cursor effects, zooms, and captions. It activates for screencast requests and rerenders after interface changes.

MIT ★ 541 updated 2026-11-09

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add dotslashsimran/simcraft --skill scripted-screencast -a claude-code
Open on GitHub

What scripted-screencast does

The skill builds a demo video from steps such as clicking, typing, and scrolling. A renderer drives the app in a headless browser, captures frames, then adds the cursor, zoom, window frame, and captions. It can also produce a vertical cut.

Store the script as demo.mjs or .json and preview it before the final render. Setup uses npm i and Playwright; video encoding requires ffmpeg on PATH or the FFMPEG variable. Use fictional brands and data: real customer information and credentials must not be recorded.

What it helps with

  • record an app walkthrough from scripted browser steps
  • add cursor movement, zoom, and captions to a demo
  • rerender a screencast after a UI change
  • create a vertical version of a product video

Example prompts for the agent

  • Record a demo of the app from the steps in demo.mjs and run a preview.
  • Make a product walkthrough with captions and a vertical cut.
  • Rerender the demo after the interface changes and show which script step failed.

How to install scripted-screencast

1Skills CLI (Vercel)

npx skills add dotslashsimran/simcraft --skill scripted-screencast -a claude-code

2Manually: put the files in place

~/.claude/skills/scripted-screencast/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/scripted-screencast/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about scripted-screencast

How do I install scripted-screencast in Claude Code?

Run npx skills add dotslashsimran/simcraft --skill scripted-screencast -a claude-code. The skill file is installed at ~/.claude/skills/scripted-screencast/SKILL.md.

What does scripted-screencast do?

It turns scripted browser actions into a demo recording with cursor effects, zoom, a window frame, and captions. The same script can render the video again after the interface changes.

What does scripted-screencast need to run?

Install dependencies with npm i; the setup uses Playwright. Video encoding needs ffmpeg on PATH or the FFMPEG variable.

Source text

SKILL.md · MIT
---
name: scripted-screencast
description: Turn a browser product demo written as a script (click, type, hover, scroll, zoom, caption) into a polished Screen Studio style recording. You get a smooth human-like cursor, click ripples, automatic zoom on small targets, typing at human speed, a padded background with a rounded window frame, burned-in captions and an optional 9:16 version that follows the cursor. It is frame-perfect and needs no retakes, and you can re-render it in minutes after the UI changes. Use when the user says "record a demo", "screen recording", "product walkthrough video", "Screen Studio style", "demo video of my app", "make a screencast", or wants to re-record a demo after a UI change.
---

# Scripted screencast

You write the demo as a list of steps. The renderer drives the app in a headless browser on a virtual clock and takes one screenshot per frame. It then adds the cursor, the zoom, the frame and the captions, and encodes the video. Every run of the same script gives the same video.

Setup, once: `cd ~/simcraft && npm i` (Playwright), and ffmpeg on the PATH or in `FFMPEG`.

## Steps

1. **Stage the app first.** Run it locally and fill it with believable data under a fictional brand: real-looking names, a day already half done, no empty states and no "test123". Put the staging in the script's `setup` (seeded localStorage, an API call, a login with a demo account) so every render starts from the same state. Never type real credentials into any site.
2. **Write the script** next to the project as `demo.mjs` (or `.json`). Start from [examples/tidy-demo.mjs](examples/tidy-demo.mjs). Use one caption per idea and keep the whole thing between 20 and 60 seconds.
3. **Preview.** This runs at 24 fps and half size, and takes about 30 seconds:
   ```
   node ~/simcraft/skills/video/scripted-screencast/scripts/render.mjs demo.mjs --preview
   ```
   Then check the zoom and framing at a few moments, and look at the stills:
   ```
   node .../render.mjs demo.mjs --preview --stills 1,4.5,9
   ```
   If a step fails, the error names the step and saves `failed-step-N.png` showing the page at that moment.
4. **Final.** This renders 60 fps at 2x pixels into a 1920x1080 video. Add `--vertical` for a 1080x1920 version as well:
   ```
   node .../render.mjs demo.mjs --vertical
   ```
5. **Report** what was staged (data, fake brand, anything faked) and give the output paths.

## Script

```js
export default {
  url: "http://localhost:3000",          // or a file path relative to the script
  viewport: { width: 1280, height: 720 }, // 1280x720 keeps text big enough for phones
  look: { background: "peach", title: "Tidy" },
  setup: [ { localStorage: { "tasks": [...] } }, { reload: true } ],  // not recorded
  steps: [
    { caption: "Add a task in seconds" },
    { click: "#newBtn" },
    { type: "Sign the gallery contract", into: "#title" },
    { click: { role: "button", name: "High" } },
    { press: "Enter" },
    { wait: ".toast" },
    { scroll: 400 },
    { zoom: ".chart" }, { pause: 1500 }, { zoom: "off" },
  ],
};
```

Targets can be a Playwright selector (`"#id"`, `"text=Save"`) or `{ role, name }`, `{ text }`, `{ label }` or `{ placeholder }`. Prefer role and visible text, because they survive redesigns better than CSS classes. The full step list, every option and all the look settings are in [reference.md](reference.md).

## Pacing rules

- Cursor moves take 400 to 900 ms, depending on the distance. This is automatic. Only set `moveMs` for a deliberate slow move.
- Hold about 600 ms after each result so people can read it. This is automatic after click, type and press. Use `hold` on a step for a bigger moment, such as 1200 ms after the main payoff.
- Zoom only when the target is small. Auto zoom does this for targets under 4% of the screen and stays in while actions are close together. Use an explicit `zoom` for a result you want people to read, and `zoom: "off"` before a wide view.
- Give the opening a second of stillness. The default `head` does this.
- One caption per idea, under 8 words, shown before the action it describes.

## After a UI change

Run the same command again. The script finds elements by role and text, so most changes need no edits. If a step fails, fix only that target (the error names it and saves a screenshot), preview, then do the final render. To change only the look (background, zoom, captions) without capturing again, render the final with `--keep` once, then use `--compose-only`.

## Rules

- Fictional brands and data only. Never record real customer data, real credentials or someone else's product presented as yours.
- Always preview before the final render and look at the stills yourself.
- Don't fake speed with `pace`: if the app is slow, stage the result or `wait` for it on camera.

© dotslashsimran · MIT · dotslashsimran/simcraft

Similar skills

SKILL.md skills catalog →

From the same set24

The whole set →
video-script The video-script Claude Code skill drafts short talking-head, explainer, and product-demo videos. It activates for requests to write or revise a script, hook, or YouTube, TikTok, Reels, or LinkedIn video. CX Codex +1 ★ 541repo stars repo-timelapse The repo-timelapse Claude Code skill turns a repository’s Git history into a video of the project changing over time. It runs on requests to show project growth or animate commits. CX Codex +1 ★ 541repo stars skill-router The skill-router Claude Code skill selects a matching skill or builds a sequence for a task. It kicks in when a request is vague or could call for several skills. CX Codex +1 ★ 541repo stars bisect-anything The bisect-anything Claude Code skill tracks a regression to a commit, dependency version, or config value. It activates for requests to find what broke a test or when a problem began. CX Codex +1 ★ 541repo stars security-threat-model The security-threat-model Claude Code skill maps threats from an app’s code and tests the highest risks locally. It activates for requests about STRIDE, app security reviews, or vulnerabilities before launch. CX Codex +1 ★ 541repo stars vox-explainer The vox-explainer Claude Code skill makes Vox-style explainer graphics or a faceless video. It runs on requests for voiceover-led visuals, animated quotes, or stats. CX Codex +1 ★ 541repo stars benchmark-honestly The benchmark-honestly Claude Code skill checks performance measurements and compares code versions. It activates for benchmark requests, load tests, latency percentiles, or claims that a change made code faster. CX Codex +1 ★ 541repo stars flaky-test-killer The flaky-test-killer Claude Code skill tracks down flaky-test causes and helps fix them at the root. It stresses the test and compares failure rates before and after the change. CX Codex +1 ★ 541repo stars

Project sponsors

Companies that keep this analytics open