Why Build Skills Instead of Agents?

TL;DR

Skills provide agents with the necessary domain expertise and procedural knowledge to perform real-world tasks effectively. Unlike traditional agents, skills are organized collections of files that can be easily shared and modified, enhancing the agent's capabilities across various domains. This approach allows for a more scalable and composable system, enabling agents to adapt and learn continuously.

Transcript

All right, good morning and thank you for having us again. Last time we were here, we're still figuring out what an agent even is. Today, many of us are using agents on a daily basis. But we still notice gaps. We still have slots, right? Agents have intelligence and capabilities, but not always expertise that we need for real work. I'm Barry. This ... Read More

Key Insights

• Skills are organized collections of files that package procedural knowledge for agents.
• Agents often lack domain expertise, which skills can provide effectively.
• Skills are simple, shareable, and can be versioned, making them accessible for both humans and agents.
• The skills ecosystem is rapidly growing, with thousands of skills being developed across various domains.
• Skills allow agents to perform complex tasks by orchestrating workflows with external data and connectivity.
• Non-technical people can create skills, making agents more accessible for everyday tasks.
• Skills can be progressively disclosed, protecting the context window while allowing for composability.
• The future of skills involves testing, evaluation, and better tooling to ensure quality and relevance.

Questions & Answers

Q: How do skills enhance an agent's capabilities?

Skills enhance an agent's capabilities by providing the necessary domain expertise and procedural knowledge to perform specific tasks. They are organized collections of files that can be easily shared, modified, and versioned, allowing agents to dynamically load and execute tasks across various domains. This approach makes agents more adaptable and effective in real-world applications.

Q: What are the different types of skills mentioned?

The different types of skills mentioned include foundational skills, third-party skills, and enterprise-specific skills. Foundational skills provide general or domain-specific capabilities that agents did not have before. Third-party skills are developed by partners to enhance agent functionality with their software. Enterprise-specific skills are tailored to teach agents organizational best practices and unique internal software usage.

Q: How do skills contribute to continuous learning for agents?

Skills contribute to continuous learning for agents by providing a standardized format for procedural knowledge that can be reused and modified over time. As agents interact with users and receive feedback, they can create new skills or update existing ones, allowing them to adapt and improve their performance. This process enables agents to learn from experience and become more effective in their tasks.

Q: Why are skills considered more scalable than traditional agents?

Skills are considered more scalable than traditional agents because they offer a modular and composable approach to enhancing agent capabilities. By packaging procedural knowledge into easily shareable and modifiable files, skills allow agents to dynamically load and execute tasks across various domains without the need for building separate agents for each use case. This makes the system more efficient and adaptable.

Q: How do skills protect the context window in agents?

Skills protect the context window in agents by being progressively disclosed. At runtime, only the metadata of a skill is shown to the model, indicating its availability. When an agent needs to use a skill, it can access the full instructions and directory, minimizing the context window usage while allowing the agent to perform tasks efficiently. This approach enables agents to handle multiple skills without overwhelming their context.

Q: What role do non-technical users play in the skills ecosystem?

Non-technical users play a significant role in the skills ecosystem by creating skills that extend agent capabilities in everyday tasks. This democratization allows a broader range of people to contribute to the development of agent functionalities, making them more accessible and relevant to various fields such as finance, recruiting, accounting, and legal. It validates the idea that skills can help non-coders enhance general agents.

Q: What trends are emerging in the skills ecosystem?

Emerging trends in the skills ecosystem include increasing complexity of skills, with some packaging software, executables, binaries, and more. Skills are complementing existing MCP servers by orchestrating workflows with external data. There's also a growing involvement of non-technical users in skill creation, validating the idea that skills make agents more accessible and adaptable for various tasks.

Q: How do skills compare to traditional software in terms of development and maintenance?

Skills are similar to traditional software in terms of development and maintenance as they are becoming more complex, requiring weeks or months to build and maintain. Like software, skills need testing, evaluation, and versioning to ensure quality and relevance. This approach treats skills as modular components that can be updated and shared, enhancing the agent's capabilities over time and making them more reliable and effective.

Summary & Key Takeaways

• Skills are designed to provide agents with domain-specific expertise, allowing them to perform tasks with greater efficiency and accuracy. This new paradigm shifts the focus from building agents to equipping them with reusable skills that enhance their capabilities across various domains.

• The skills ecosystem is growing rapidly, with foundational, third-party, and enterprise-specific skills being developed. These skills enable agents to perform tasks such as document creation, scientific research, and browser automation, making them more effective in different environments.

• Skills are accessible to non-technical users, allowing for a broader range of people to contribute to the development of agent capabilities. This democratization of skill creation helps agents become more adaptable and useful in real-world applications, leading to continuous learning and improvement.