Subtitle Packager

Takes a normalized transcript JSON with stable timing data and converts it into actual subtitle files like SRT or ASS. You'd reach for this when transcription is already done and you just need proper chunking, timing normalization, and clean export packaging for editors or distribution. The skill is strict about its boundaries: it won't call speech-to-text models, won't guess at missing timing data, and will stop cold if required inputs aren't there. It keeps intermediate work under .postplus and puts final deliverables where you'd expect them. Honest take: the discipline around not inventing timing or bypassing errors is refreshing, even if it means you occasionally get a hard stop instead of a helpful guess.