Save Hours Every Day with an AI Skill

Get back a few hours every day without giving up good videos? That sounds like a marketing promise. It is simply a TV guide - built for YouTube, with artificial intelligence.

I built an AI skill that analyzes YouTube videos for me before I watch them. The skill gives me an elevator pitch, the gap in the argument, and the conclusion - all tailored to exactly what I care about. Instead of sinking into hours of video, I decide in one minute what is worth it.

The TV guide for YouTube

People used to browse the TV guide and know whether tonight's movie was worth their time. Nothing like that exists for YouTube: thumbnails and titles lure you in without naming the actual benefit. This skill closes exactly that gap - with a summary based on my own criteria, for any video I choose.

The problem: without a preview, you pay with lifetime

The creator playbook is simple: a promising title, a curiosity-driven thumbnail - and in the end a video that stretches a single statement across twenty minutes. Time is money, and anyone who has to stay up to date professionally cannot afford bad purchases at scale. The original version of this problem is old: the classic TV guide existed precisely because airtime was precious.

The obvious fix - reading summaries online - has a catch: they judge by the criteria of whoever wrote them, not by mine. What bores one person may be exactly the core I need. That is why this calls for a made-to-measure solution.

What the skill delivers

Elevator pitch

Three to four sentences: What do you learn in this video? Why is it worth it? If those questions are not convincingly answered after reading, there is no need to watch.

The gap

Why does the topic seem complicated? What does the author leave out? What is promised but never resolved? This gap in the argument often decides whether you end up disappointed - or whether it simply does not matter to you.

Practice box

The key concepts of the video in short form plus a concrete scenario for daily work. That lets you judge immediately whether the content applies to your own project.

Deep analysis

A technical assessment: architecture fit, scalability, complexity, relevance for current projects. No full movie, just the conclusion - and the reasoning behind it.

How the technology works

The skill is deliberately simple. It downloads the video's subtitles with the open-source tool yt-dlp - the video itself never needs to be downloaded. From that transcript, the AI agent then generates the structured report as an HTML file: with a table of contents, video metadata, and the four analysis sections.

Everything runs locally in your own agent - for example Claude Code. No cloud service, no subscription, no extra sign-up. Once you understand how a skill is structured, you can adapt the criteria to your own questions at any time: different sections, different audience, different language.

Use the skill for free

The complete instructions are available here as a downloadable Markdown file. Drop the file into your agent's skills folder as SKILL.md - after that, a single call with the YouTube URL is enough. The instructions are written in German and work with any agent that can run Bash.

Download the skill instructions (.md)

The conclusion

The real gain is not this one skill, but the pattern behind it: for almost every recurring task you can build a made-to-measure skill that answers exactly the questions that matter to you. Once you know your criteria, you no longer need other people's summaries - and you only spend your time where it returns something.

➡️ All content in the embedded video was created with AI.