Export an indexed transcript

Export only material already present in the local corpus. This workflow is deterministic and must not invoke an LLM or acquire missing material.

Workflow

1. Resolve the requested video through the read-only MCP. If MCP is unavailable and the request does not already contain an unambiguous YouTube URL or video ID, ask for one instead of searching a cwd-relative catalog. Stop unless one existing video is identified.
2. Use the language requested by the user. Otherwise omit --lang; ask the user only if the MCP evidence already shows multiple languages or the export command returns the explicit language choices.
3. Run yt-insights export video VIDEO_OR_URL --format FORMAT --lang LANGUAGE --json, omitting --lang when the transcript is unambiguous. Supported formats are md, txt, and vtt.
4. Return the exact output path, video_id, language, format, and source_sha256 from the JSON result.

Do not overwrite an existing file unless the user explicitly authorizes --force. Do not acquire a missing video, edit the transcript, invent a source hash, or treat an output path as proof before the command succeeds.