What problem does it solve? Turning an online video link into a clean, readable transcript usually means downloading the video or running speech recognition. This Skill reads the subtitle tracks the platform already serves (YouTube, Bilibili, or generic sites), prefers human tracks over auto-generated ones, removes speech noise, and delivers a complete article in the user's language. ## Core Features & Use Cases - Platform-agnostic pipeline: Verified adapters for YouTube (timedtext capture, transcript panel) and Bilibili (view/dm-view APIs), plus generic discovery via network sniffing, textTracks, and <track> tags for any other site. - Multiple access backends: Pure HTTP/CLI, agent-browser page injection, Tampermonkey userscript, or WebBridge evaluate against the user's logged-in browser, with lawful-use confirmation before full export. - Editorial delivery contract: Captures all timed cues with coverage verification, removes filler words and false starts while preserving every substantive claim, then restructures and fully translates into the user's language before summaries or Q&A. - Use Case: A user pastes a Bilibili or YouTube link and asks for the content as an article; the agent confirms lawful use, pulls the complete human subtitle track, verifies cue coverage, and returns a polished translated article. ## Quick Start Give the agent a video link and ask it to extract the subtitles and deliver a complete article in your language, confirming lawful use when prompted.
