December 19, 2025 Anthropic is expanding a set of enterprise tools designed to make workplace AI more consistent, manageable and reusable, as businesses struggle to get real productivity gains from chatbots that aren’t connected to how people actually work. The company has upgraded “skills” in Claude and is releasing Agent Skills as an open standard, a bet that portability and ecosystem adoption will matter as much as model quality.
Skills are reusable instruction sets that package an organisation’s workflows, standards and domain knowledge so users don’t have to rebuild complex prompts from scratch each time. In practice, that can look like brand and writing guidelines, repeatable email templates or task creation workflows that plug into tools such as Jira and Asana. Anthropic says the latest updates make skills easier to build, discover and deploy across teams, with central management for enterprise admins.
The company is also leaning into integrations. Skills now connects with common workplace platforms including Notion, Canva, Figma and Atlassian, and Anthropic has launched a directory of partner-built skills that also includes vendors such as Stripe and Zapier. The idea is to make AI work less like a standalone chat window and more like a layer that can reliably operate inside the tools employees already use.
Anthropic’s bigger move is to turn Agent Skills into an open standard. “We’re launching Agent Skills as an independent open standard with a specification and reference SDK,” Anthropic product manager Mahesh Murag said in an interview, adding that Microsoft has adopted Agent Skills within VS Code and GitHub and that the company is in discussions with other ecosystem players. Anthropic argues that standardisation makes skills portable across different models and platforms. Hence, a workflow built for Claude can also run elsewhere if those products adopt the standard.
Under the hood, Anthropic says skills are designed to be lightweight. Each skill can be summarised in a small number of tokens, with full detail loaded only when needed, a technique intended to let organisations build large libraries without overwhelming a model’s working context.
Anthropic is framing skills as a way to standardize how AI performs work inside companies. Employees can describe what they want in plain language, and Claude can generate a skill that encodes those steps for repeated use. According to Anthropic, enterprise customers are already using skills “in production” across coding and business functions such as legal, finance, accounting and data science. The company added that its public repository has attracted significant community traction.
The timing of the expansion coincides with a shift in AI design. Instead of building dozens of specialized agents, companies are moving toward one general assistant equipped with a growing library of capabilities. That model simplifies governance, reduces operational overhead and makes AI easier to audit and improve over time.
