Anthropic launches enterprise ‘Agent Skills’ and opens the standard, challenging OpenAI in workplace AI
Anthropic said Wednesday it would release it Agent skills technology as an open standard, a strategic bet that sharing the approach to making AI assistants more capable will strengthen the company’s position in the rapidly evolving enterprise software market.
The San Francisco-based artificial intelligence company also unveiled organization-wide management tools for enterprise clients and a list of partner-developed enterprise skills, including Atlassian, Figma, Kanva, Stripe, ConceptAnd Zapier.
These steps mark a significant expansion of a technology that Anthropic first introduced in October, transforming what started as a niche developer role into infrastructure that is now on the cusp of becoming an industry standard.
“We are launching Agent Skills as an independent open standard with a specification and reference SDK available at https://agentskills.io” said Mahesh Murag, product manager at Anthropic, in an interview with VentureBeat. “Microsoft has already adopted Agent Skills within VS Code and GitHub; so do popular coding tools like Cursor, Goose, Amp, OpenCode and more. We are having active conversations with others across the ecosystem.”
Within the technology that teaches AI assistants to do specialized work
Skills are essentially folders of instructions, scripts, and resources that tell AI systems how to perform specific tasks consistently. Rather than requiring users to create extensive prompts every time they want an AI assistant to complete a specialized task, skills package procedural knowledge into reusable modules.
The concept addresses a fundamental limitation of large language models: while they have broad general knowledge, they often lack the specific procedural expertise needed for specialized professional work. For example, a skill for creating PowerPoint presentations might include preferred formatting conventions, slide structure guidelines, and quality standards—information that the AI loads only when working on presentations.
Anthropic designed the system around what it calls “progressive disclosure.” Each skill costs only a few dozen tokens when summarized in the AI’s context window, with full details loaded only when the task requires them. This architectural choice allows organizations to deploy extensive skill libraries without overwhelming the AI’s working memory.
Fortune 500 companies already use legal, financial and accounting skills
The new business management features enable Anthropic’s administrators to work Team And Enterprise plans to deliver skills centrally, controlling which workflows are available in their organizations while allowing individual employees to customize their experience.
“Enterprise customers use skills in manufacturing for both coding workflows and business functions such as legal, finance, accounting and data science,” said Murag. “The feedback was positive because their skills allowed them to personalize Claude to the way they actually work and achieve high-quality output faster.”
The community response has exceeded expectations, according to Murag: “Our skills repository has already surpassed 20,000 stars on GitHub, with tens of thousands of community-created and shared skills.”
Atlassian, Figma, Stripe, and Zapier will be added to Anthropic’s skill list at launch
Anthropic is launching with the skills of ten partners, a roster that reads like a who’s who of modern enterprise software. The presence of Atlassianwhat makes Jira And Confluencein addition to design tools Figma And Kanvapayment infrastructure company Stripeand automation platform Zapier suggests that Anthropic positions Skills as the connective tissue between Claude and the applications companies are already using.
The business arrangements with these partners are focused on ecosystem development rather than immediate revenue generation.
“Partners building capabilities for the directory do so to improve the way Claude works with their platforms. It is a mutually beneficial ecosystem relationship similar to MCP connector partnerships,” Murag explains. “There are no revenue sharing arrangements in place at this time.”
Anthropic takes a measured approach when vetting new partners. “We started with established partners and are developing more formal criteria as we expand,” said Murag. “We want to create a valuable skillset for businesses while helping partners’ products shine.”
It is notable that Anthropic does not charge extra for the option. “Skills work on all Claude surfaces: Claude.ai, Claude Code, the Claude Agent SDK, and the API. They are included at no additional cost in Max, Pro, Team, and Enterprise plans. API usage follows standard API pricing,” said Murag.
Why Anthropic is giving away its competitive advantage to OpenAI and Google
The decision to release Skills as an open standard is a calculated strategic choice. By making skills transferable across AI platforms, Anthropic is betting that ecosystem growth will benefit the company more than proprietary lock-in.
The strategy seems to be working. OpenAI has quietly adopted a structurally identical architecture in both ChatGPT and are Codex CLI tool. Developer Elias Judin discovered the implementation earlier this month and found folders containing skill files that mirror Anthropic’s specifications: same file naming conventions, same metadata format, same directory organization.
This convergence suggests the industry has found a common answer to a tough question: How do you ensure AI assistants are consistently good at specialized work without expensive model tuning?
The timing aligns with broader standardization efforts in the AI industry. Anthropic donated its Model Context Protocol to the Linux Foundation on December 9, and both Anthropic and OpenAI co-founded the Agentic AI Foundation next to Block. Google, Microsoft and Amazon Web Services have joined. The foundation will manage several open specifications, and Skills fits naturally into this drive for standardization.
“We have also seen how complementary skills and MCP servers are,” said Murag. “MCP provides secure connectivity to external software and data, while skills provide the procedural knowledge to use these tools effectively. Partners who have invested in strong MCP integrations were a natural starting point.”
The AI industry is abandoning specialized agents in favor of one assistant that learns everything
The Skills approach is a philosophical shift in the way the AI industry thinks about making AI assistants more capable. The traditional approach involved building specialized agents for different use cases: a customer service agent, a encryption agent, a research agent. Skills suggest a different model: one general-purpose agent equipped with a library of specialized capabilities.
“We used to think that agents in different domains would look very different,” Barry Zhang, an anthropic researcher, said at an industry conference last month, according to a Business Insider report. “The agent underneath is actually more universal than we thought.”
This insight has significant implications for enterprise software development. Rather than building and maintaining multiple specialized AI systems, organizations can invest in creating and managing capabilities that encode their institutional knowledge and best practices.
Anthropic’s own internal investigation supports this approach. A study released by the company in early December found that engineers used Claude in 60% of their work, achieving a self-reported 50% improvement in productivity – a two- to threefold increase over the previous year. Notably, 27% of the work Claude supported consisted of tasks that would otherwise not have been accomplished, including building internal tools, creating documentation, and addressing what employees called “papercuts” – small quality of life improvements that had been consistently deprioritized.
Security risks and skills atrophy are emerging as concerns for enterprise AI implementations
The Skills framework is not without possible complications. As AI systems become more capable through skills, questions arise about retaining human expertise. Anthropic’s internal research found that while skills enabled engineers to work in more domains (backend developers building user interfaces, researchers creating data visualizations), some employees were concerned about skills atrophy.
“When producing output is so easy and fast, it becomes increasingly difficult to actually take the time to learn,” said an Anthropic engineer in the company’s internal survey.
There are also safety considerations. Skills provide Claude with new capabilities through instructions and code, meaning malicious skills could theoretically introduce vulnerabilities. Anthropic recommends only installing skills from trusted sources and thoroughly checking those from less trusted sources.
The open standard approach also introduces management questions. Although Anthropic has published the specification and launched a reference SDK, the long-term management of the standard remains undefined. Whether it will fall under the Agentic AI Foundation or need its own governance structure is an open question.
Anthropic’s real product may not be Claude; it can be the infrastructure that everyone else builds on
Skills’ trajectory reveals something important about Anthropic’s ambitions. Two months ago, the company introduced a feature that looked like a developer tool. Today, that feature has become a specification that Microsoft builds into VS Code, which OpenAI replicates in ChatGPT, and which major software giants are eager to support.
The pattern reflects strategies that have previously reshaped the technology industry. Companies from Red Hat to Google have discovered that open standards can be more valuable than proprietary technology; that the company that defines how an industry works often generates more value than the company that tries to own it directly.
For enterprise technology leaders evaluating AI investments, the message is clear: skills become infrastructure. The expertise that organizations encode into skills today will determine how effectively their AI assistants perform tomorrow, regardless of which model powers them.
The competition between Anthropic, OpenAI and Google will continue. But to the question of how AI assistants can become reliably good at specialized work, the industry has quietly arrived at an answer — and it came from the company that gave it away.
