# RESOURCE2SKILL: Transforming Videos into Agent Skills > Source: > Published: 2026-07-15 08:56:15+00:00 # RESOURCE2SKILL: Transforming Videos into Agent Skills RESOURCE2SKILL taps into the potential of tutorial videos and multimodal resources to create executable skills for software agents. This approach not only enhances agent performance but also revolutionizes how we think about skill acquisition. In a world where AI is rapidly evolving, RESOURCE2SKILL steps up by doing what others largely overlook, turning the vast repository of tutorial videos and [multimodal](/glossary/multimodal) content into actionable skills for software agents. If you've ever wondered how to make better use of those endless YouTube tutorials, this framework might just have the answer. ## A New Kind of Skill Library For too long, skill libraries for AI agents have been confined to the predictable: hand-written guides, text-heavy manuals, or agent-specific traces. These methods, while useful, miss out on the rich, detailed knowledge found in videos, articles, and other media. RESOURCE2SKILL changes the game by distilling these resources into what can be thought of as a multimodal Skill Wiki. Each entry in this Skill Wiki isn't just a stale line of code. It's a vibrant mix: structured text, executable code, visual examples, and even metadata. Think of it as a one-stop shop for everything an agent might need to execute a skill. Videos bring the temporal and visual elements. Code offers executable patterns. Articles and artifacts add depth and style. ## Real-World Impact Why does this matter? Simple. The framework improves agents' performance significantly. In seven practical authoring domains, RESOURCE2SKILL boosted average scores by 11.9 percentage points over agents without skills. It even outperformed strong baselines in 26 out of 28 model-domain tests. That's not just an improvement. it's a transformation. Every model that runs offline is a vote for private computing. The framework organizes skills hierarchically and allows online acquisition when existing skills aren't enough. It proves that with the right setup, AI agents can learn and adapt in real-time, effectively getting smarter with each interaction. ## Utility Over Hype But what makes RESOURCE2SKILL stand out is its embrace of source diversity and selection strategy. It doesn't just pull from the same old sources. Instead, it leverages a mix of repository types, ensuring that each skill is grounded and adaptable. Utility, not hype. That's the point. So, what's holding us back from adopting this kind of framework universally? It's a question worth pondering as we continue to push the boundaries of AI. As more developers and researchers embrace this multimodal approach, the potential for AI agents becomes nearly limitless. On-device AI isn't coming. It's here. And RESOURCE2SKILL is a testament to how a fresh perspective can redefine what's possible in AI skill acquisition. Get AI news in your inbox Daily digest of what matters in AI.