agentleFS
ParadevAIParadevAI/Paradev/.agents/skills/mobile-demo-film/SKILL.md

mobile-demo-film

Skill16k starsChanged 14 days agoMITView on GitHub

Turn mobile screen recordings into a configurable demo video with phone footage, title cards, cutaways, and an end card. Use when asked to edit a mobile feature demo or launch clip. Use mobile-demo-record to capture real app footage first.

  • Installs packages

1 other project uses a version of this; 1 new project in the last 4 weeks, against 0 the 4 before.

Adopted by

Surging: 1 new project took it in the last 4 weeks, none the 4 weeks before.thin data

Week of Sep 20: 0 new projectsWeek of Sep 27: 0 new projectsWeek of Oct 4: 0 new projectsWeek of Oct 11: 0 new projectsWeek of Oct 18: 0 new projectsWeek of Oct 25: 0 new projectsWeek of Nov 1: 1 new projectWeek of Nov 8: 0 new projects
Sep 2026Oct 2026Nov 2026
New projects taking a version of this file each week, over the 8 weeks our latest run read: only 1 in all, so this is a hint, not a trend.

Most recently taken up by

Who's talking about ParadevAI/Paradev

The biggest 5 of 53 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 53

Talked about on GitHub, Bluesky and YouTube (53 times).

On GitHub · 50

45 more on GitHub

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill to edit a new demo from real app recordings.

    line 8
  2. Very specific:

    Use Python 3.10+, Pillow 10.1+, and ffmpeg (including ffprobe and libx264).

    line 15
  3. Distinctive:

    Use that environment's Python for the commands below.

    line 25
  4. Distinctive:

    Plan one story: Write a short table of on-screen actions and durations.

    line 31
  5. Very specific:

    Add --frames 12 --height 480 to customize sampling and thumbnail height.

    line 45
  6. Distinctive:

    Keep typing and reading at a natural pace.

    line 51
  7. Very specific:

    Add a render object to the spec.

    line 80
  8. Distinctive:

    Keep cut points within the source duration and speed positive.

    line 120
  9. Very specific:

    Run film.py --help or a subcommand with --help for CLI options.

    line 129
  10. Distinctive:

    After editing the renderer, run its focused tests (requires ffmpeg and Pillow):

    line 134

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Mobile demo film

Use this skill to edit a new demo from real app recordings. The bundled renderer produces silent video with configurable dimensions, frame rate, framing, and fades. Defaults are 1080x1920 at 30fps, with a phone frame on a warm paper backdrop. A useful default is one feature in 20 to 40 seconds; adapt the story and pacing to the request.

Prerequisites

Use Python 3.10+, Pillow 10.1+, and ffmpeg (including ffprobe and libx264). Recording with mobile-demo-record needs macOS; editing existing footage does not. Font discovery tries macOS system fonts, then Linux DejaVu fonts, then Pillow’s fallback. Specify font files for consistent typography across machines. Install Pillow in a scratch virtual environment:

python3 -m venv <scratch>/film-venv
<scratch>/film-venv/bin/pip install "Pillow>=10.1"

Use that environment's Python for the commands below. Commands assume the repo root as working directory. Output directories are created automatically. Existing outputs are overwritten; source assets and the composition spec cannot be used as the film output.

Workflow

  1. Plan one story. Write a short table of on-screen actions and durations. Start with the user's action, show the app responding, and end on the visible result. A tour can start with the finished artifact instead. Verify the feature works before filming it.
  2. Capture real footage with mobile-demo-record. Use demo data and a separate take for each beat. If the story needs an agent doing work, use a working host and agent. Capture desktop cutaways with cdp-verification when needed.
  3. Normalize before choosing cuts. Simulator recordings have variable frame rates: python3 .agents/skills/mobile-demo-film/scripts/film.py normalize raw/01.mov cfr/01.mp4. Choose start/end timestamps from this constant-rate copy. Add --fps 24 for a 24fps project (the default is 30).
  4. Compose using a JSON spec beside the recordings: python3 .agents/skills/mobile-demo-film/scripts/film.py compose <scratch>/film.json.
  5. Review a contact sheet and play the entire film: python3 .agents/skills/mobile-demo-film/scripts/film.py sheet <scratch>/film.mp4 <scratch>/sheet.jpg. Add --frames 12 --height 480 to customize sampling and thumbnail height. By default, the sheet samples eight frames; inspect transitions and typing in playback too. Check for clipped text, unreadable UI, private information, spinners, and unintended clock changes.
  6. Deliver the film and identify the build, environment, staged data, and omitted steps. Draft accompanying post copy only if requested; publishing needs user authorization.

Keep typing and reading at a natural pace. Trim or speed up waiting, and make time jumps clear. Use title cards to explain missing context rather than implying an unrecorded action happened in the app. Framing, masks, and the simulated Dynamic Island are presentation layers; do not fabricate app state inside the recording.

Composition spec

This example combines phone footage, a desktop cutaway, a result, and an end card. Replace its paths, cut points, and text for the feature being demonstrated.

{
  "output": "film.mp4",
  "backdrop": "#F2F0EB",
  "segments": [
    { "type": "phone", "src": "cfr/01.mp4", "start": 1, "end": 7, "label": "Start on your phone" },
    { "type": "card", "eyebrow": "Meanwhile", "title": "The agent works on your Mac.",
      "image": "shots/desktop.png", "duration": 4 },
    { "type": "phone", "src": "cfr/02.mp4", "start": 0, "end": 10 },
    { "type": "end", "wordmark": "Paradev", "caption": "Review the result anywhere", "duration": 2.2 }
  ]
}

Media, output, and optional titleFont/sansFont paths resolve relative to the spec file; absolute paths also work. Top-level backdrop, bezel, ink, and muted accept #RRGGBB.

Output and framing

Add a render object to the spec. Existing specs without it keep the portrait defaults. For a landscape demo:

"render": {
  "width": 1920,
  "height": 1080,
  "fps": 24,
  "crf": 18,
  "fade": 0.2,
  "phone_width": 0.45,
  "phone_height": 0.8,
  "bezel_width": 0.012,
  "corner_radius": 0.1
}

For a square clip, use "width": 1080, "height": 1080. Cards and typography scale with the canvas; the centered phone fits both width and height limits while preserving source aspect ratio. Use raw for unframed desktop footage.

Render optionMeaning and default
width, heightCanvas pixels; even integers at least 64. Defaults: 1080, 1920
fpsInteger 1–120; default 30
crfFinal H.264 quality, 0–51; lower means higher quality/larger files. Default 18
fadeFade duration in seconds; 0 disables fades. Default 0.35; capped at half each framed/still segment’s duration; raw clips have no fade
phone_width, phone_heightMaximum screen width/height as fractions of canvas dimensions, greater than 0 and at most 0.95. Defaults: 0.615, 0.82
bezel_widthBorder thickness as a fraction of the smaller canvas dimension, 0–0.25. Default 0.015
corner_radiusScreen corner radius as a fraction of screen width, 0–0.25. Default 0.137

Segments

SegmentFields
phonesrc, optional start/end in seconds, speed (default 1), label, island (false disables the overlay)
rawsrc, optional start/end, speed; unframed footage fitted to the output canvas with black padding
cardtitle, optional eyebrow, caption, image, footer, duration (default 4 seconds)
endwordmark or transparent PNG logo, optional caption, duration (default 2.2 seconds)

Keep cut points within the source duration and speed positive. Every segment must last at least one output frame. The renderer validates all segments and source paths before encoding. Card titles wrap; captions, labels, and footers should be short. The renderer removes audio, so use another editing workflow if narration or app audio is needed.

Store raw takes, specs, and finished films in a scratch directory. Share the inputs and spec when teammates need to re-edit a particular demo; they are not required to create a new one.

Reusing and checking the renderer

Run film.py --help or a subcommand with --help for CLI options. The Python module can also be imported: normalize(src, dst, fps=30), compose(spec_path) (returns the output Path), and sheet(src, dst, frames=8, height=900) work without invoking the CLI. RenderSettings holds layout calculations without mutable global canvas settings.

After editing the renderer, run its focused tests (requires ffmpeg and Pillow):

python3 -B -m unittest discover -s .agents/skills/mobile-demo-film/scripts -p 'test_*.py'

These render small portrait, landscape, and square compositions, check output dimensions, frame rate and duration, and exercise invalid specs. Also inspect a contact sheet at the intended delivery size; automated checks do not judge readability or composition.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.