video-to-text-agentlisted

# Video To Text Agent Use this skill when a user provides a local video and asks to understand its narration, cover text, selling points, script, or other short-video content. The production tool is the Enterprise repository's non-interactive CLI: ```bash "${ENTERPRISE_AI_OFFICE_ROOT}/scripts/transcribe" ``` The legacy personal wrapper and interactive command were not migrated. Do not invoke, source, or depend on them during Agent execution. This skill does not use OCR. Spoken narration is the authoritative content source, and the source video filename stem is the cover text. ## Path Rules For a source video: ```text <media-file>/pre-coated with 3M tape,Mass potting line.mp4 ``` the derived values are: ```text cover_text: pre-coated with 3M tape,Mass potting line result_dir: the Enterprise runtime transcript output for the source basename transcript: the Enterprise runtime transcript output for the source basename ``` The result directory must be beside the video and have exactly the same name as the video filename without its final extension. Do not add `_文案` or any other suffix. The cover text is the exact filename stem. Do not translate, rewrite, or otherwise alter it when reporting the extracted source information. ## Agent Commands For Chinese or Chinese-heavy narration, use SenseVoice: ```bash "${ENTERPRISE_AI_OFFICE_ROOT}/scripts/transcribe" "/absolute/path/video.mp4" \ --engine sensevoice --language zh ``` For English or mixed-language narration, use