Run3 video keyframe extraction

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill run3_video-keyframe-extraction

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Use this skill to extract keyframes (I-frames) from an MP4 video file using ffmpeg. This produces numbered PNG files in a specified output directory.

1.3 KB, as published. Nobody here has run it

Use ffmpeg to extract only I-frames (keyframes) from the video:

ffmpeg -i /root/super-mario.mp4 -vf "select=eq(pict_type\,I)" -vsync vfp /root/keyframes_%03d.png -y

This will produce files like /root/keyframes_001.png, /root/keyframes_002.png, etc.

After extraction, list the files to confirm how many keyframes were produced:

ls -1 /root/keyframes_*.png | sort

Count them:

ls -1 /root/keyframes_*.png | wc -l

The output format keyframes_%03d.png is zero-padded to 3 digits, starting from 001.
ffmpeg numbers sequentially starting at 1.
All keyframes are in timeline order by default.
If the number of keyframes seems wrong, you can also try extracting at a fixed interval (e.g., 1 fps) using -vf fps=1 instead, but I-frame extraction is the standard approach for "key frames."
Use ffprobe to check total number of I-frames beforehand if needed:

ffprobe -select_streams v -show_frames -show_entries frame=pict_type /root/super-mario.mp4 2>/dev/null | grep "pict_type=I" | wc -l