sticker-frames

name	sticker-frames
description	Extract individual PNG frames from any WhatsApp sticker — static .webp, animated .webp, or Lottie .was. Use when the user wants to see, analyze, clone, or recreate a sticker.
allowed-tools	Bash(extract-frames:*),Bash(ffmpeg:*),Bash(ffprobe:*)

Sticker frame extraction

WhatsApp stickers arrive as files in /workspace/attachments/:

Extension	Format	What it is
.webp	RIFF/WebP	Static image OR animated WebP
.was	WhatsApp Animated Sticker	Lottie JSON (usually gzip-wrapped) — vector animation

You cannot view a .was file directly — it is vector data, not raster. This skill rasterizes both formats into a numbered sequence of PNG frames you can Read with vision.

Usage

extract-frames /workspace/attachments/sticker-1777730095530.was
extract-frames /workspace/attachments/sticker-1777730095530.webp /tmp/my-frames

• First arg: path to the sticker file.
• Second arg (optional): output directory. Defaults to /tmp/sticker-frames-<nanos>/.
• Stdout: newline-separated list of frame PNG paths, in order.
• Exit code: 0 on success, non-zero with stderr message on failure.

Cap is 60 frames by default (covers typical sticker length). For longer animations pass --max-frames N after the path.

What to do with the frames

This skill does not auto-send to the user — frames are an input for your analysis, not output for the chat.

After extraction, the typical flow is:

1. Read each frame PNG (vision).
2. Describe what you see — motion, palette, style, timing.
3. If the user wants to clone or recreate it, recommend a representation (SVG keyframes, Canvas animation, a new Lottie JSON, CSS keyframes) based on what the animation actually does.

If the user just sent a sticker without asking anything specific, ask what they want to do with it before extracting — extraction takes a few seconds.

Failure modes

• unknown format — magic bytes don't match WebP, gzip, or raw JSON. The file may be corrupt or a format WhatsApp introduced after this skill was written. Report the magic bytes from stderr to the user.
• Lottie render hangs — if render-lottie.mjs doesn't finish in ~30s, the JSON may be invalid or use a Lottie feature lottie-web doesn't support. Read the JSON header (head -c 500 /tmp/.../lottie.json) and report.
• ffmpeg single-frame output for animated WebP — some animated WebPs only expose one frame to the demuxer. If you only get frame_001.png for what looks animated, that's the source's fault, not the skill's.