Short Video Parser

This package implements exactly the parsing and transcription behavior it advertises, but the registry metadata is incomplete and there are privacy implications you should consider before installing: - The transcribe flow uploads extracted audio to SiliconFlow (Bearer token from .env or --api_key). If you provide that API key the skill will send your audio to a third‑party service. If you don't want audio uploaded, do not provide the API key or avoid running the transcription step. - SKILL.md and scripts require ffmpeg on PATH (used via subprocess). The registry does not declare this requirement — install ffmpeg if you plan to transcribe. - The code downloads videos and stores them in tmp/ and writes Markdown reports to demos/. If you enable auto_cleanup=false or omit cleanup, media files are retained on disk. - The optional parse_api_url allows using an external parsing API; setting it causes your share URLs to be sent to that endpoint — only set it to a URL you trust. - Packaging mismatch: registry metadata not declaring required env vars/binaries is a red flag (sloppy or incomplete packaging). Prefer to inspect/run the code in a sandbox or VM first. Review .env and scripts/transcribe.py to confirm endpoints and remove or replace the external transcription endpoint if needed. Recommended actions before use: inspect the code locally, run in an isolated environment, avoid supplying API keys you don't trust, and consider building a local/no‑upload transcription path if you need to protect sensitive content.