Anthropic said Wednesday he would publish his own Agent skills technology as an open standard, a strategic assumption that sharing its approach to improving the performance of AI assistants will strengthen the company’s position in the rapidly growing enterprise software market.
The San Francisco-based artificial intelligence company also showcased organization-wide management tools for enterprise clients and a catalog of partner-developed skills from companies including: Atlassian, Figma, Canva, Stripe, ConceptAND Zapier.
The moves mark a significant expansion of the technology that Anthropic first introduced in October, transforming what was initially a niche development function into an infrastructure that now appears to be an industry standard.
“We are launching Agent Skills as an independent open standard with the specification and reference SDK available at https://agentskills.io” said Mahesh Murag, product manager at Anthropic, in an interview with VentureBeat. “Microsoft has already implemented agent skills in VS Code and GitHub; as well as popular coding agents like Cursor, Goose, Amp, OpenCode and others. We are in active conversations with others across the ecosystem.”
Inside the technology that teaches AI assistants to perform specialized jobs
Skills are essentially folders containing instructions, scripts, and resources that tell AI systems how to consistently perform specific tasks. Instead of requiring users to create complicated prompts every time they want the AI assistant to perform a specialized task, skills can aggregate procedural knowledge into reusable modules.
This concept recognizes a fundamental limitation of gigantic language models: although they have broad general knowledge, they often lack the specialized procedural knowledge needed for specialized professional work. For example, the ability to create presentations in PowerPoint may include preferred formatting conventions, slide structure guidelines, and quality standards – information that the AI only loads as you work on presentations.
Anthropic designed the system around what it calls “progressive disclosure.” Each skill only requires a few dozen tokens when summarized in the AI context window, and full details only load when the task requires them. This architectural choice enables organizations to deploy extensive skill libraries without taxing the AI’s working memory.
Fortune 500 companies are already leveraging skills in law, finance and accounting
Up-to-date enterprise management features enable Anthropic’s administrators Team AND Undertaking plans to centrally deliver skills by controlling which workflows are available in their organizations while enabling individual employees to customize their experience.
“Enterprise customers leverage manufacturing skills both in coding processes and in business functions such as legal, finance, accounting and data analytics,” Murag said. “The response was positive because the skills allowed them to adapt Claude to how it actually works and produce high-quality prints faster.”
According to Murag, the community’s response has exceeded expectations: “Our skills repository has already surpassed 20,000 stars on GitHub and includes tens of thousands of skills created and shared by the community.”
Atlassian, Figma, Stripe and Zapier join Anthropic’s skills catalog at launch
Anthropic launches with the skills of ten partners, a list that reads like a who’s who of state-of-the-art enterprise software. Presence Atlassianwhat makes Yes AND Fugitivealong with design tools Figma AND Canvapayment infrastructure company Stripeand automation platform Zapier suggests that Anthropic positions Skills as the connective tissue between Claude and the applications companies already utilize.
Business arrangements with these partners focus on ecosystem development rather than immediate revenue generation.
“Partners who develop catalog skills do so to improve Claude’s interoperability with their platforms. It is a mutually beneficial ecosystem relationship, similar to an MCP connector partnership,” Murag explained. “There are no revenue sharing arrangements at this time.”
Anthropic takes a measured approach when vetting novel partners. “We started with established partners and as we grow, we are developing more formal criteria,” Murag said. “We want to create a valuable skillset for enterprises while helping partner products shine.”
It’s worth noting that Anthropic does not charge extra for this opportunity. “Skills work on all Claude platforms: Claude.ai, Claude Code, Claude Agent SDK and API. They are included in the Max, Pro, Team and Enterprise plans at no additional cost. API use is subject to standard API pricing,” Murag said.
Why Anthropic is giving up its competitive advantage to OpenAI and Google
Decision to dismiss Skills as an open standard it is a calculated strategic choice. With the ability to transfer skills to AI platforms, Anthropic anticipates that developing an ecosystem will bring greater benefits to the company than being locked into ownership.
The strategy seems to be working. OpenAI has quietly adopted a structurally identical architecture in both cases ChatGPT and him Codex CLI tool. Developer Elias Judin discovered this implementation earlier this month by finding directories containing skill files matching the Anthropic specification – same file naming conventions, same metadata format, and same directory organization.
This convergence suggests that the industry has found a common answer to a vexing question: How do we make AI assistants consistently good at specialized jobs without pricey model tuning?
The timeline is consistent with broader standardization efforts in the AI industry. Anthropic given its Model Context Protocol to the Linux Foundation on December 9, and both Anthropic and OpenAI are co-founders Agent AI Foundation next to Block. Google, Microsoft and Amazon Web Services have joined as members. The Foundation will manage many open specifications, and skills will naturally fit into this focus on standardization.
“We also saw how MCP’s skills and servers complement each other,” Murag noted. “MCP provides secure connectivity to external software and data, and the skills provide the procedural knowledge necessary to effectively use these tools. Partners who have invested in strong MCP integration were a natural starting point.”
The AI industry is abandoning specialized agents in favor of a single assistant that learns everything
The Skills approach is a philosophical shift in the way the AI industry thinks about enhancing the capabilities of AI assistants. The established approach was to build specialized agents for different utilize cases – customer service agent, coding agent, research agent. Skills suggest a different model: one general-purpose agent equipped with a library of specialized capabilities.
“We used to think that agents from different domains would look very different,” anthropic researcher Barry Zhang said at an industry conference last month, according to Business Insider report. “The agent underneath is actually more versatile than we thought.”
This knowledge has critical implications for enterprise software development. Instead of building and maintaining multiple specialized AI systems, organizations can invest in creating and refining skills that encode their institutional knowledge and best practices.
Property of Anthropic internal research supports this approach. A study released by the company in early December found that engineers were using Claude for 60% of their work, achieving a self-reported 50% enhance in productivity, representing a two- to three-fold enhance in productivity compared to the previous year. It’s worth noting that 27% of the work Claude did involved tasks that otherwise wouldn’t have been done, including building internal tools, creating documentation, and dealing with what employees call “paper cuts,” compact quality of life improvements that were consistently deprioritized.
Security threats and skills attrition are emerging as concerns around AI implementations in enterprises
The Skills framework is not without potential complications. As AI systems become more skill-based, questions arise about maintaining human knowledge. Anthropic’s internal research found that while skills had enabled engineers to work in more domains – back-end developers creating user interfaces, researchers creating data visualizations – some employees were worried about skills atrophy.
“When it’s so easy and fast to produce results, it becomes increasingly difficult to devote time to learning,” one Anthropic engineer said in an internal company survey.
There are also safety considerations. Skills provide Claude with novel capabilities through instructions and code, which means malicious skills could theoretically introduce security vulnerabilities. Anthropic recommends only installing skills from trusted sources and carefully reviewing skills from less trusted sources.
The open standard the approach also introduces management issues. Although Anthropic has published the specification and launched a reference SDK, the long-term management of the standard remains undetermined. The open question is whether it will fall under the Agentic AI Foundation or require its own governance structure.
Anthropic’s real product may not be Claude – it may be the infrastructure that everyone else relies on
The Skills trajectory reveals something critical about Anthropic’s ambitions. Two months ago, the company introduced a feature that looked like a developer tool. Today, this feature has become a specification that Microsoft builds into VS Code, that OpenAI replicates in ChatGPT, and that enterprise software giants strive to support.
This pattern reflects strategies that have previously transformed the technology industry. Companies from Red Hat to Google have discovered that open standards can be more valuable than proprietary technology – that a company that defines how an industry works often provides more value than a company that tries to take ownership of it.
For enterprise technology leaders evaluating AI investments, the message is plain: skills become infrastructure. The expertise that organizations encode in skills today will determine how effectively their AI assistants will perform tomorrow, regardless of the model that supports them.
The competitive battles between Anthropic, OpenAI and Google will continue. But when it comes to how to make AI assistants reliably good at specialized jobs, the industry has quietly found an answer, and it came from the company that provided it.