human launched a new feature on Thursday. Claude Eye This marks the company’s latest effort to make artificial intelligence more practical for enterprise workflows as it chases rival OpenAI in the growing race to develop AI-powered software.
A feature called skillallows users to create folders containing instructions, code scripts, and reference materials that Claude can automatically load when relevant to a task. This system represents a fundamental shift in how organizations customize their AI assistants, moving beyond one-time prompts to reusable packages of domain expertise that work consistently across the enterprise.
"Skills is based on the belief and vision that as model intelligence continues to improve, we will continue to move toward general-purpose agents that have access to their own file systems and computing environments." Mahesh Murag, a member of Anthropic’s technical staff, said in an exclusive interview with VentureBeat. "Agents initially only know the name and description of each available skill, and can choose to load more information about a particular skill if it’s relevant to the task at hand."
This announcement was made as Anthropic and was rated as follows. $183 billion in recent $13 billion funding rounda recent report predicts that annual revenue could almost triple to $26 billion in 2026. Reuters report. The company is now nearing $7 billion in annual revenue, up from $5 billion in August, largely due to enterprise adoption of its AI coding tools, a market it faces stiff competition from OpenAI’s recently upgraded Codex platform.
How “progressive disclosure” solves the context window problem
skill This is fundamentally different from existing approaches to customizing AI assistants, such as prompt engineering and search augmentation generation (RAG), Murag explained. The architecture relies on what Anthropic calls "progressive disclosure" — Claude initially sees only the skill name and short description, then autonomously decides which skill to load based on the task at hand, accessing only the specific files and information he needs at that moment.
"Unlike RAG, this relies on simple tools that allow Claude to manage the file system and read files from the file system." Murag told VentureBeat. "A skill can contain an unlimited amount of context to teach Claude how to complete a task or set of tasks. This is because the skill is based on the premise that agents can autonomously and intelligently navigate the file system and execute code."
This approach allows organizations to bundle much more information than traditional context windows allow, while maintaining the speed and efficiency that enterprise users demand. A skill can include step-by-step instructions, code templates, reference documentation, brand guidelines, compliance checklists, and executable scripts. All of this is organized into a folder structure that Claude navigates intelligently.
The configurability of this system provides another technical advantage. Multiple skills are automatically stacked together when required for complex workflows. For example, Claude calls on his corporate brand guidelines skills, financial reporting skills, and presentation formatting skills simultaneously to generate quarterly investor materials and coordinates all three without manual intervention.
How the skill differs from OpenAI’s Custom GPT and Microsoft’s Copilot
Anthropic positions skills as different from competitors like OpenAI Custom GPT and Microsoft’s Copilot StudioHowever, these features address similar enterprise needs for AI customization and consistency.
"Skills’ combination of gradual disclosure, composability, and bundling of executable code is unique on the market." Murag said. "While other platforms require developers to build custom scaffolds, Skills allows anyone, technical or non-technical, to create a professional agent by organizing their procedural knowledge into files."
Cross-platform portability also makes your skills stand out. The same skills work the same in any country claude eye, claude code (Anthropic’s AI coding environment), the company’s API,and Claude Agent SDK For building custom AI agents. Organizations can develop skills once and deploy them everywhere teams use Claude, a huge benefit for companies looking for consistency.
This feature supports any programming language compatible with the underlying container environment, and Anthropic provides a sandbox for security. However, the company acknowledges that users will need to carefully scrutinize which skills they trust before allowing AI to run their code.
Early customers report an 8x increase in productivity in their financial workflows
Early customer implementations reveal how organizations are adapting skill Automate complex knowledge tasks. Japanese e-commerce giant RakutenAI teams are using their skills to transform financial operations that previously required manual coordination across multiple departments.
"Skills streamline management accounting and finance workflows," Yusuke Kaji, general manager of AI at Rakuten, said in a statement. "Claude uses our procedures to process multiple spreadsheets, detect significant anomalies, and generate reports. What once took a day can now be completed in an hour."
That’s an 8x increase in productivity for certain workflows. This is the kind of measurable return on investment that companies are increasingly looking for from their AI implementations. Mike Krieger, chief product officer at Anthropic and co-founder of Instagram, recently pointed out that companies are moving past the past. "AI FOMO" Specific success indicators are required.
design platform Kamba plans to integrate the skills into its own AI agent workflows. "Canva plans to leverage skills to customize agents and expand what they can do." Anwar Hanif, general manager and head of ecosystem at Canva, said in a statement: "This unlocks new ways to embed Canva even deeper into agent workflows, empowering teams to capture their unique context and easily create engaging, high-quality designs."
cloud storage provider box sees skills as a way to make enterprise content repositories more practical. "Skills teach Claude how to interact with Box content." Yashodha Bhavnani, Head of AI at Box, said: "Users can save hours of effort by converting saved files into PowerPoint presentations, Excel spreadsheets, and Word documents that adhere to organizational standards."
Corporate security question: Who controls which AI skills are available to employees?
For corporate IT departments, skill This raises important questions about governance and control, especially since this feature allows AI to execute arbitrary code in a sandboxed environment. Anthropic has built administrative controls that allow enterprise customers to manage access at the organizational level.
"Enterprise administrators can control access to skill functionality through administrative settings, enable or disable access, and monitor usage patterns." Murag said. "Once enabled at the organization level, individual users must still opt in."
This two-tier consent model (organizational enablement and individual opt-in) reflects lessons learned from previous enterprise AI deployments, where bulk deployments raised compliance concerns. However, Anthropic’s governance tools appear to be more limited than some enterprise customers might expect. The company currently does not provide granular control over the specific skills employees can use or a detailed audit trail of custom skill content.
Organizations concerned about data security should note that skills require a cloud code execution environment that runs in isolated containers. Anthropic advises users: "Stick to trusted sources" Although it provides security documentation when installing skills, the company recognizes that this is an inherently riskier feature than traditional AI interactions.
From API to no-code: How Anthropic makes skills accessible to everyone
Anthropic takes several approaches. skill Accessible by users with a wide range of technical knowledge. For non-technical users claude eyeprovided by the company "skill creator" Skills that interactively guide users to build new skills by asking workflow questions and automatically generating folder structures and documentation.
Developers we work with Anthropic’s API You can gain programmatic control through the new /skills endpoint and manage skill versions through the Claude Console web interface. To use this feature, you must enable Code Execution Tools Beta in your API requests. If you are a Claude Code user, you can install skills through a plugin from the Anthropics/skills GitHub marketplace, and teams can share skills through a version control system.
"Skills are included in Max, Pro, Teams, and Enterprise plans at no additional charge." Murag confirmed. "API usage is subject to standard API pricing." This means that your organization only pays for the tokens consumed while the skill is running, not the skill itself.
Anthropic offers several pre-built skills for common business tasks, including professionally creating Excel spreadsheets with formulas, PowerPoint presentations, Word documents, and fillable PDFs. These human-created skills will continue to be free.
Why skill initiation is important in the AI coding war with OpenAI
The skill announcement comes at a pivotal time in Anthropic’s competition with OpenAI, particularly in AI-assisted software development. Anthropic launched just one day before Skills Claude Haiku 4.5comparable in coding performance despite being a smaller and cheaper model. claude sonnet 4 — which was state-of-the-art when it was released just five months ago.
This rapid improvement curve reflects the breakneck pace of AI development, where today’s frontier capabilities become tomorrow’s commodities. OpenAI is also focused on its coding tools and recently upgraded its tools. codex platform and GPT-5 and expand GitHub co-pilot ability.
Anthropic’s revenue trajectory — it could be reached. $26 billion in 2026 An estimated $9 billion by the end of 2025, suggesting the company is successful in converting corporate profits into paying customers. The timing also follows this week’s announcement that Salesforce is deepening its AI partnerships with both OpenAI and Anthropic to power its Agentforce platform, suggesting companies are taking a multi-vendor approach rather than standardizing on a single provider.
Skills address real-world problems. "rapid engineering" The problem is that effective use of AI relies on individual employees creating elaborate instructions for daily tasks, with no way to share that expertise across teams. Skills transform tacit knowledge into explicit, shareable assets. For startups and developers, this feature can significantly accelerate product development and add advanced documentation generation capabilities that previously required a dedicated engineering team and weeks of development.
The composability aspect suggests a future in which organizations build libraries of specialized skills that can be combined to suit increasingly complex workflows. Pharmaceutical companies may develop skills in regulatory compliance, clinical trial analysis, molecular modeling, and patient data privacy to work together seamlessly to create customized AI assistants with deep expertise across multiple disciplines.
Anthropic has indicated that it is working on simplified skill creation workflows and enterprise-wide deployment capabilities to make it easier for organizations to distribute skills across large teams. As this capability is rolled out to Anthropic’s more than 300,000 enterprise customers, the real test will be whether organizations find the skills to be materially more useful than their existing customization approaches.
So far, Skills is the clearest expression yet of Anthropic’s vision for AI agents. That is, rather than a generalist who tries to do everything rationally, it is an intelligent system that knows when to access specialized knowledge and can coordinate multiple knowledge areas to accomplish complex tasks. Once that vision takes hold, the question will no longer be whether your company uses AI, but whether the AI knows how your company actually functions.
