Anthropic releases Agent Skills as open standard, challenging OpenAI in enterprise AI race

Anthropic Makes Agent Skills Open Standard for Cross-Platform Use

Anthropic announced Wednesday it would release Agent Skills as an open standard, marking a strategic shift that could position the company as a dominant AI infrastructure layer across the industry 4

4

. The specification and reference SDK are now available at agentskills.io, with Microsoft already adopting Agent Skills within VS Code and GitHub 2

2

. Popular coding agents including Cursor, Goose, Amp, and OpenCode have also integrated the technology, with Anthropic reporting active conversations with others across the ecosystem 4

4

.

The open standard approach mirrors Anthropic's earlier success with the Model Context Protocol (MCP), which became the de facto communication standard for AI agents to connect with third-party tools and data systems 5

5

. By making Agent Skills portable across different AI platforms, the company is betting that ecosystem growth will deliver more value than proprietary lock-in, potentially allowing skills created in Claude to work with models like ChatGPT or other platforms that adopt the standard 3

3

.

Enterprise AI Tools Enable Organization-Wide Management

The updates include new organization-wide management capabilities that allow Team and Enterprise administrators to provision skills centrally, controlling which workflows are available across their organizations while letting individual employees customize their experience 4

4

. This centralized approach addresses a fundamental challenge in workplace AI adoption, where scattered tools have contributed to messy and fraught deployment experiences 3

3

.

Mahesh Murag, product manager at Anthropic, explained that enterprise customers are using skills in production across coding workflows and business functions like legal, finance, accounting, and data science. "The feedback has been positive because skills let them personalize Claude to how they actually work and get to high-quality output faster," Murag said in an interview 4

4

. The company's skills repository has already crossed 20,000 stars on GitHub, with tens of thousands of community-created and shared skills 4

4

.

Partner Skills Directory Expands Workplace AI Capabilities

Anthropic launched with a partner skills directory featuring pre-built skills from ten major companies including Atlassian, Figma, Canva, Stripe, Notion, and Zapier 1

1

. These reusable instruction sets enable AI agents to perform specialized tasks consistently, such as applying brand style guidelines to marketing materials, creating tasks in Jira and Asana, and automating email templates 3

3

. The roster represents connective tissue between Claude and applications businesses already use, with partners building skills to enhance how Claude works with their platforms through mutually beneficial ecosystem relationships 4

4

.

Skills are folders containing instructions, scripts, and resources that tell large language models how to perform specific tasks consistently, eliminating the need for users to craft elaborate prompts each time 4

4

. The system uses "progressive disclosure," where each skill takes only a few dozen tokens when summarized, with full details loading only when tasks require them 4

4

. Skills are included in Max, Pro, Team, and Enterprise plans at no additional cost, with API usage following standard pricing 4

4

.

Challenging OpenAI in Enterprise AI Implementation

While the consumer race remains dominated by ChatGPT and Google's Gemini, Anthropic's models continue to outpace both OpenAI and Google in the enterprise market 3

3

. The company has distinguished itself by staying focused on enterprise customers with offerings designed to make working professionals' lives easier 1

1

. OpenAI has quietly adopted structurally identical architecture in both ChatGPT and its Codex CLI tool, with developers discovering implementations that mirror Anthropic's specification 4

4

.

Murag noted the complementary nature of Skills and MCP: "MCP provides secure connectivity to external software and data, while skills provide the procedural knowledge for using those tools effectively" 2

2

. Chatbots can't achieve meaningful productivity gains and return on investment for businesses until they're tailored to how users actually work, making this enterprise AI implementation critical for adoption 3

3

. Skills now reside in the Tools sidebar with a quick-create flow where users simply describe what they want and Claude builds it 1

1

.