Pulls specific frames from local video files based on what you actually need them for, not just uniform sampling. If you're doing persona research, it prioritizes faces. Product visibility gets object-focused extraction. UI walkthroughs grab text-heavy frames. The mode defaults are smart: hook analysis pulls from the first few seconds, CTA checks grab the end, shot breakdowns trigger on scene changes. It's designed to feed downstream work, so outputs include contact sheets, timestamps, and selection reasoning. The boundary discipline is strict: it won't guess your intent or fake its way past CLI errors, which means less cleanup later when you're actually using these frames.
npx skills add https://github.com/postplusai/postplus-skills --skill frame-extraction