Agent skill

Launch Film

by SkylarKitchen in SkylarKitchen/skills

Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than…

MITAuto-check passedGame Development

Install Launch Film

skills CLI
$ npx skills add SkylarKitchen/skills --skill launch-film -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install SkylarKitchen/skills launch-film --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/SkylarKitchen/skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/design/launch-film .claude/skills/launch-film && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
launch-film
GitHub stars
117
Token cost
~1.9k tokens
SKILL.md length
1,172 words
Files
14
Skills in repo
3
Repo updated
First seen
Licence
MIT

At a glance

Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than…

  • Works in 7 steps: Source the surface. node… → Write plan.js in the project (copy… → Render: node engine/capture.cjs plan.js… → …
  • The user wants a launch video
  • SKILL.md covers Mode: pick one first, Length and rhythm, Pipeline and Sound, plus 2 more sections
  • Runs JavaScript, Shell and Python scripts from its folder; calls node, bash and ffmpeg

What it does

Launch Film is an agent skill from SkylarKitchen/skills. Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than cut from a screen recording. Use when the user wants a launch video, a teaser, a "better version" of their own screen recording, or a showcase clip of something they built. Looping background reels cut from existing footage are a different job.

Its SKILL.md is about 1.9k tokens, which your agent loads only when the skill is triggered. The skill folder holds 18 other files (for example `agents/openai.yaml`, `engine/sound.py` and `examples/quickstart/plan.js`).

It sits in Game Development, covering 3D graphics and WebGL. It works with Three.js. The repository describes itself as: Agent skills for design work: 3D build guides and more. The licence is MIT.

When your agent uses it

  • The user wants a launch video
  • A better version of their own screen recording
  • A showcase clip of something they built

Example prompts

  • “better version”
  • “/launch-film”

Requirements

  • Python 3
  • Node.js
  • A Bash shell

Workflow steps

7 steps, taken from the first numbered list in SKILL.md.

  1. Source the surface. node engine/shoot.cjs page.png 1440 900 (add full for a scrolling page). For 3D, render the frame sequences first (see…
  2. Write plan.js in the project (copy examples/quickstart/plan.js). See the plan reference below.
  3. Render: node engine/capture.cjs plan.js preview.mp4. It needs Playwright and ffmpeg; CHROMIUM= uses a system Chromium.
  4. Look at it. Make a sheet with ffmpeg -i preview.mp4 -vf "fps=2,scale=320:-1,tile=8x5" -frames:v 1 sheet.png, and pull full-size frames mid…
  5. Gate. bash measure.sh preview.mp4 must pass: after the opener, calm shots have a median of at least 3.5 s, with at most two rapid bursts…
  6. Show the user the preview and the sheet before any final encode.
  7. Deliver. Re-render with CRF=18 PRESET=slow. X takes H.264 1920×1080. For the feed, a 4:5 cut (frame: [864, 1080], scale: 1.25 = 1080×1350)…

What it can do on your machine

Read from SKILL.md and the folder at commit fb53232. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships script files (JavaScript, Shell and Python), which the agent can run.

    Shell commands in SKILL.md call:

    • node
    • bash
    • ffmpeg

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Launch Film loads about 1.9k tokens when it runs. Until then it costs about 114 tokens; SKILL.md has 1,172 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~114
When it runs · the whole SKILL.md, loaded when a task matches
~1.9k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from SkylarKitchen/skills at commit fb53232, republished under its MIT licence (© SkylarKitchen). 1,172 words, ~1,889 tokens.

Download SKILL.mdSave it as .claude/skills/launch-film/SKILL.md (or your agent's skills folder). This skill also uses 13 other files; get the full folder from GitHub.
name
launch-film
description
Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than cut from a screen recording. Use when the user wants a launch video, a teaser, a "better version" of their own screen recording, or a showcase clip of something they built. Looping background reels cut from existing footage are a different job.

Launch film

The films are authored, not cut. Energy comes from a camera moving over one continuous surface, with eased push-ins on the details that matter, not from edits. Every frame is a pure function of time, so a render is repeatable and a 25 s film at 1080p renders in under a minute.

The engine is engine/film.html, a 2D canvas that plays a plan.js. engine/capture.cjs steps it frame by frame in headless Chromium and pipes the frames to ffmpeg.

Mode: pick one first

  • Showcase (the default): the work speaks. The opener is an accelerating flip through the project's own images (photos, drawings, stills), with no words. The body is one continuous camera over the real surface, with push-ins on the best details. It ends on the pulled-back product, or on a plain title if the user wants one. Ask before adding a title card. Add one plain caption per beat at most (people watch X muted); no taglines, no pitch copy.
  • Interaction: the cursor acts on the real page. A click motivates each camera move, and hovers show the page's real states. Shoot cursor-free stills of each state with Playwright (hover() and click() draw no cursor), plus each element's centre. Then use the product scene's states (section swaps keyed by time), hovers (a band, underline or ring where the page has no hover style), cursor.clicks and cursor.hide.
  • 3D product (a Three.js app): headless Chromium has no GPU, so render the 3D shots as frame sequences on a real GPU in a normal browser. A debug hook drives the camera and app state per frame, and each frame is POSTed as a JPEG to a local sink. Render at 4K, downscale to 1080p with lanczos, and play them in frames scenes. from: {image, wipe} wipes a photo into the render from the same camera. The UI take is the real app driven headless with real pointer events, one screenshot per frame (SwiftShader draws WebGL at about 1 s a frame), with the pointer recorded to cursor.json and replayed through a product scene with seq and cursor.frames. Dry-run each drag first and read the app's own state after it: a move that looks fine can turn a piece or break the app's checks. Worked example: examples/room-configurator/, a film of a room-from-photos page (film.js renders the shots, ui.cjs records the UI take, plan.js and plan45.js hold the timeline).

Length and rhythm

  • Teaser, 20–30 s for X: opener about 2 s, body in two or three beats of 4–8 s each.
  • Feature, about 80 s: a few continuous takes with push-ins. Calm shots have a median of 4–8 s.
  • Hard cuts only between scenes. Rapid cuts live only in the opener, never over UI states. Longer than 90 s is a tutorial, not this skill.

Pipeline

  1. Source the surface. node engine/shoot.cjs <url> page.png 1440 900 (add full for a scrolling page). For 3D, render the frame sequences first (see the worked example). Measure every position you'll push in on from the full-size image, never from a scaled contact sheet.
  2. Write plan.js in the project (copy examples/quickstart/plan.js). See the plan reference below.
  3. Render: node engine/capture.cjs plan.js preview.mp4. It needs Playwright and ffmpeg; CHROMIUM=<path> uses a system Chromium.
  4. Look at it. Make a sheet with ffmpeg -i preview.mp4 -vf "fps=2,scale=320:-1,tile=8x5" -frames:v 1 sheet.png, and pull full-size frames mid camera move. The sheet hides direction errors, such as a panel dropping in from above instead of rising.
  5. Gate. bash measure.sh preview.mp4 <opener_end_s> must pass: after the opener, calm shots have a median of at least 3.5 s, with at most two rapid bursts of up to 5 s. Then run bash holds.sh preview.mp4 <t> ... at each camera hold. It fails a push-in that frames blank page.
  6. Show the user the preview and the sheet before any final encode.
  7. Deliver. Re-render with CRF=18 PRESET=slow. X takes H.264 1920×1080. For the feed, a 4:5 cut (frame: [864, 1080], scale: 1.25 = 1080×1350) takes more of a phone screen. Give it its own plan that imports the 16:9 one and overrides the frame, the camera keys and the type sizes. A 3D film needs its shots re-rendered at 4:5: a narrower frame keeps the photo's height, not its width.
Show full SKILL.md (466 more words)Show less

Sound

Deliver X cuts without an audio track (-an) unless the user supplies music or asks for sound: X autoplays muted, and generic synth beds read as cheap. capture.cjs writes <out>.cues.json with every cut, camera whip and press. engine/sound.py (needs numpy) turns the cues into a synth bed if the user wants one. With supplied music, snap scene durations to its beats.

Plan reference

Top level:

  • frame and scale: the logical frame (default [1280, 720]) and output scale. Positions and sizes are in logical px.
  • fps: default 25. Use 30 to match frame sequences rendered at 30, and keep every duration a whole number of frames (n / 30).
  • images: {name: url}, preloaded.
  • fonts: {Name: url} or {Name: {src, weight: "100 900"}} for a variable face; font picks the one to use. Always load one. A minimal Linux container may map sans-serif to a monospace face.
  • captions: labels in film time, {text, from, to, row, tag}. They sit top-left, in sentence case: a white label, with an optional dark letter tag such as an option's A, B or C. caption sets x, y, gap, size, small, ink and bg.
  • grain: true: adds film grain.

Scenes ({type, dur, bg, ...}), played in order:

  • still: a project image. cover: true makes it full-bleed with a slow push; otherwise fit, dx and dy place it. word sets optional text.
  • product: a UI image floating on the ground. crop: [x, y, w, h] of the image; camera: [[t, [cx, cy, zoom]], ...], eased between keys, in image px from the crop's centre (positive y looks lower on the page). seq swaps in a screenshot sequence instead of image. cursor is either {path: [[t, [x, y]]], click, clicks, hide} or {frames, fps, scale, ripple} (the pointer recorded with the take, in CSS px times scale). Also states, hovers and headline (a line over the empty ground before the UI rises in).
  • frames: a full-bleed frame sequence, seq: {src: "shot_%04d.jpg", count, fps, at}. It holds the first frame before at and the last frame after the end. from: {image, wipe: [t0, t1]} wipes it in over a still.
  • title: a centred text with an optional sub line, rising in; typeOn: true types the text instead.
  • tagline: lines rising in, one after another.

Gotchas

  • A whip (a camera jump over 400 image px between keys) adds a whoosh cue. Any other camera change stays silent.
  • The scene detector behind measure.sh misses a cut between two same-colour frames, so check a white-on-white opener by eye.
  • In holds.sh, test camera holds only. A title card is meant to read as empty.
  • Check every photo before it goes public: documents with readable text, house numbers, street signs, and names on screens.
  • The capture browser has no GPU. Keep film scenes 2D canvas, and bring 3D in as pre-rendered frames.

© SkylarKitchen, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 13 other files in skills/design/launch-film of SkylarKitchen/skills.

  • SKILL.md
  • agents/openai.yaml
  • engine/capture.cjs
  • engine/film.html
  • engine/shoot.cjs
  • engine/sound.py
  • examples/quickstart/plan.js
  • examples/room-configurator/README.md
  • examples/room-configurator/film.js
  • examples/room-configurator/plan.js
  • examples/room-configurator/plan45.js
  • examples/room-configurator/ui.cjs
  • holds.sh
  • measure.sh

Open the folder on GitHubat commit fb53232

Compare with similar skills

Launch Film next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Launch Film compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Launch Film this skillSkylarKitchen/skills117—~1.9kAutomated safety check: PassMIT
Image to Three.js Modelimg2threejs/img2threejs18k1 repos~8.2kAutomated safety check: PassApache-2.0
Web CloneJane-xiaoer/claude-skill-web-clone1k1 repos~2.7kAutomated safety check: PassMIT
Threejs Game Directormajidmanzarpour/threejs-game-skills2.5k—~2.2kAutomated safety check: PassMIT
Threejs Gameplay Systemsvalkor-ai/loom1.2k1 repos~1.4kAutomated safety check: PassApache-2.0
Threejs World Generationcalesthio/OpenMontage66k—~2kAutomated safety check: PassAGPL-3.0

Similar skills

  • Image to Three.js Model

    img2threejs/img2threejs

    Rebuilds the object in a reference image as a procedural, animation-ready Three.js model written entirely in code, using staged sculpting with quality checks.

    18k GitHub starsUsed in 1 repo~8.2k tokens
    Game DevelopmentAuto-check passed
  • Web Clone

    Jane-xiaoer/claude-skill-web-clone

    网站复刻 / 克隆方法论。USE WHEN 用户说 复刻网站、克隆网站、clone website、抄个站、仿站、 照着这个站做一个、reproduce site、还原某个网页效果、把这个站搬下来改成我的、 复刻某个交互/WebGL/Canvas/Three.js 效果。提供「先拿真源码 → 判路径 → 逆向拆解 → 搭工程 → 替换内容」的可移植决策树,覆盖静态站 /…

    1k GitHub starsUsed in 1 repo~2.7k tokens
    Game DevelopmentAuto-check passed
  • Threejs Game Director

    majidmanzarpour/threejs-game-skills

    Entrypoint for building, upgrading, and finishing Three.js browser games.

    2.5k GitHub stars~2.2k tokensUpdated 13 days ago
    Game DevelopmentAuto-check passed
  • Build and iterate playable Three.js game systems: starter scaffold, architecture, design briefs, core loops, level and encounter design, entities, input, camera, collision and physics, scoring…

    1.2k GitHub starsUsed in 1 repo~1.4k tokens
    Game DevelopmentAuto-check passed
  • Threejs World Generation

    calesthio/OpenMontage

    Build deterministic, editable, free-viewpoint Three.js worlds from text or structured briefs.

    66k GitHub stars~2k tokensUpdated 7 days ago
    Game DevelopmentAuto-check passed
  • 3Dviz Pro Max Scene Builder

    viettranx/3dviz-pro-max

    Guides building an expressive 3D scene in Three.js or Blender by reasoning through intent, object construction and scientific grounding before using a catalog of recipes and kits.

    709 GitHub stars~2.8k tokensUpdated 1 mo ago
    Game DevelopmentAuto-check passed

More from SkylarKitchen/skills

  • 3D Assembly Manual

    SkylarKitchen/skills

    Builds a phone-first interactive 3D step-by-step build guide as one self-contained Three.js HTML file, starting from a bundled starter template.

    117 GitHub stars~1.6k tokensUpdated 13 days ago
    Auto-check passed
  • Room From Photos

    SkylarKitchen/skills

    Builds a 3D model of a real room from a few phone photos, with movable furniture layouts, as one Three.js HTML page you can open in a browser.

    117 GitHub stars~6k tokensUpdated 13 days ago
    Auto-check passed

Works with

Questions about Launch Film

What does Launch Film do?

Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than…. Launch Film is an agent skill from SkylarKitchen/skills.js scene, project photos), authored as a plan and rendered on a canvas engine rather than cut from a screen recording.

When should I use Launch Film?

Launch Film fits situations like: the user wants a launch video; A better version of their own screen recording; A showcase clip of something they built.

How do I install Launch Film in Claude Code?

Run `npx skills add SkylarKitchen/skills --skill launch-film -a claude-code`. Or copy the skill folder (skills/design/launch-film in SkylarKitchen/skills) into .claude/skills/launch-film in your project. Claude Code loads it when a task matches its description.

How do I install Launch Film in Codex?

Run `npx skills add SkylarKitchen/skills --skill launch-film -a codex`. Or copy the skill folder (skills/design/launch-film in SkylarKitchen/skills) into .agents/skills/launch-film in your project. Codex loads it when a task matches its description.

Can I use Launch Film in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add SkylarKitchen/skills --skill launch-film -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/launch-film, .gemini/skills/launch-film, .github/skills/launch-film and .opencode/skills/launch-film in your project.

What does Launch Film need to run?

Going by SKILL.md and its folder, Launch Film needs JavaScript, a shell and Python for the scripts in its folder and the command-line tools its instructions call (node, bash and ffmpeg). Our summary lists: Python 3; Node.js; A Bash shell.

Does Launch Film access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Launch Film safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Launch Film use?

Launch Film is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Launch Film use?

About 1.9k tokens (SKILL.md is roughly 7.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Launch Film?

Skills that share tags, products or a category with Launch Film: Image to Three.js Model (img2threejs/img2threejs, 18k stars), Web Clone (Jane-xiaoer/claude-skill-web-clone, 1k stars), Threejs Game Director (majidmanzarpour/threejs-game-skills, 2.5k stars) and Threejs Gameplay Systems (valkor-ai/loom, 1.2k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Launch Film?

SkylarKitchen (a GitHub user) maintains it in SkylarKitchen/skills, which has 117 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on September 28, 2026.

Source: SkylarKitchen/skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.