Paste a URL. We watch the video, extract what the expert actually does differently, and hand you:
No signup for your first video. No credit card.
Three layers — one for humans, one for agents, one for runtimes.
A 1-page markdown summary of every decision pattern the presenter demonstrated, with timestamps and tradeoffs — so you know what you're about to hand to your agent.
Seven platform adapters: Claude Code, Claude.ai Projects, ChatGPT Custom GPT, Gemini Gems, Cursor, Windsurf/Cline, raw system prompt. Copy → paste → done.
Typed action DAGs with preconditions, postconditions, and value-adherence checks. Run them locally with vidskill execute, or let your agent orchestrate them.
Most tools summarize videos. We extract the decision patterns — falsifiable claims about what the expert does differently — and compile them into skills that link every claim back to a timestamp in the source.
Pure-prompt skills that work in any LLM chat. No installs.
Skills that read and write files — perfect for Claude Code or Cursor.
Skills that run commands or drive a browser — ship with our runtime.