YouTube Video Transcript

Fetch, summarize, and save YouTube transcripts with timestamp navigation, chapter detection, and searchable content.

This skill appears coherent and implements what it advertises, but consider the following before installing: - Installation: it requires yt-dlp (Homebrew or pip). Approve installing yt-dlp from those package sources if you trust them. - Caching & privacy: transcripts are stored under ~/youtube-video-transcript/ only with your explicit consent; refuse caching if you don't want local copies. The skill promises to show and delete cached files on request — verify that behavior after use. - Cookies & auth: the skill may recommend using your browser cookies for age-restricted videos; never paste passwords or tokens into chat. Provide cookies only if you understand the implications and do not expect the skill to store them. - Batch processing risk: the skill includes batch/channel/playlist workflows (mass extraction). This is legitimate for research but can be privacy- or copyright-sensitive; don't enable bulk processing without confirming you have the right to archive that content. - Proactive behavior: it may offer to automatically transcribe when you share YouTube links — only enable that if you want the agent to act on links without an explicit per-video consent. If you want stronger assurance, test the skill on a short, non-sensitive video and confirm where files are written and what gets cached before using it at scale.

Type: OpenClaw Skill Name: youtube-video-transcript Version: 1.0.0 The skill bundle is designed for a legitimate purpose: fetching, summarizing, and saving YouTube transcripts locally using `yt-dlp`. All file system operations are confined to `~/youtube-video-transcript/` and explicitly require user consent for caching, with strong emphasis on transparency. The markdown instructions for the AI agent consistently reinforce ethical behavior, user privacy, and secure practices (e.g., 'Never ask for or store credentials', 'Cache With Consent'), actively mitigating prompt injection risks rather than exploiting them. There is no evidence of intentional data exfiltration, backdoor installation, or other malicious activities.