This is for cleaning up screen recordings when you've got long pauses and keyboard noise cluttering your video. It runs two FFmpeg scripts in sequence: first strips out silent segments using an aggressive threshold tuned for typing sounds, then speeds up what's left by 1.2x. The silence detection is pretty configurable if the defaults cut too much, and there's a dry run mode so you can preview before committing. It's opinionated about the workflow order, which makes sense since speeding up first would mess with the audio characteristics. The main limitation is that long videos take a few minutes to process since each segment gets handled individually.
npx skills add https://github.com/zc277584121/marketing-skills --skill raw-video-processing