Anthropic introduced on Thursday a suite of updates to Skills, a capability that trains Claude on tasks tailored to how users work. The update shows how the market is moving beyond large language models, themselves, and becoming more focused on use cases.
As part of the update, the AI model maker released its Agent Skills specification as an open standard. Agent Skills expands the ability for Claude users to create, deploy, share and discover new skills for agentic AI, according to Anthropic.
The vendor is adding an organization-wide management capability, which allows enterprises administrators on its Team and Enterprise plans to manage Skills in a central location. Users will also have access to prebuilt Skills from Anthropic partners including Canva, Notion, Figma and Atlassian.
The move to make Agent Skills an open standard follows Anthropic’s success with its Model Context Protocol, an open standard for large language models to communicate. The vendor donated MCP to the Linux Foundation earlier this month.
Diving Deep Into Skills
The expansion of Skills for Claude comes two months after Anthropic introduced Agent Skills, a new method for building specialized agents using files and folders. Inside a folder, users can add files to describe processes such as filling out a form or browsing a website, teaching the agent how to perform certain tasks.
“For many of the business use cases that require human input, now you can actually have an AI model with the skill go ahead and do that action on your behalf,” said Arun Chandrasekaran, an analyst at Gartner. “It’s giving more agentic capabilities to these models so that they can go and perform actions on your behalf.”
Skills also allow agents to run software without needing to talk to another tool or orchestrate between different programs, said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget.
“The agent can just resolve the task on its own, and in that case, the agent becomes a lot smarter, a lot easier to deploy,” Su said. “Obviously, that then requires a whole host of skills, related management systems and updates.”
With the newly expanded suite of Skills, Anthropic seeks to create a skills library and a set of predefined skills that can be applied across different types of tasks, Chandrasekaran said.
Concerns and Maturity
Giving an agent added capabilities means giving them more autonomy to complete specific tasks, which could raise concerns among enterprises.
However, Su said, that this is why enterprises need to have a strong AI security and governance strategy.
“Enterprises need to have guardrails and filters and firewalls in place when they are deploying all these different agents with skills,” he said.
Anthropic’s push into developing Skills is an example of how the focus of the AI market has evolved away from model updates to use cases.
“The industry is now moving beyond the models into, ‘Hey, what can the models do on my behalf? How can the models orchestrate workflows? How can models re-engineer processes? How can AI become more autonomous within our environment?'” Chandrasekaran said. He added that, aside from Anthropic, OpenAI Operator is similar to Skills because the agent is performing specific tasks. “These are all like incremental capabilities that are required for us to move toward those goals.”
