video-to-skilllisted

<!-- argument-hint: <video-url-or-local-path>... [skill-name] --> # Video-to-Skill Convert demonstrated capability, not a transcript summary. Own the complete workflow from accessible source discovery through installation of one course-specific Skill. ## User contract A normal invocation is: ```text /video-to-skill <URL> # Claude Code $video-to-skill <URL> # Codex ``` Treat supplied URLs or local paths as authorization for reversible local processing. Process every accessible playlist or course item by default, resume cached work after interruption, and keep the private evidence workspace separate from the portable Skill. Ask only when a material boundary cannot be inferred safely: credentials are needed, local or private media would leave the machine, a billed dependency is required, scope is unexpectedly large, the user must choose between materially different curricula, or an installation name conflicts with different content. Never ask the user to run internal commands, supervise extraction, select frame parameters, copy task payloads, or invoke a second tutor Skill. Never request an account password. Resolve the complete expected source set before treating acquisition as complete. Preserve playlist order and explicit complete, partial, failed, skipped, inaccessible, or retired states. Continue through isolated source failures when useful material remains, never silently omit an expected item, and report how missing material limits the generated Skill. T