Anthropic Opens Agent Skills as an Industry Standard for AI Workflows

Anthropic has made its Agent Skills specification publicly available as an open standard, extending the capabilities of its Claude platform for enterprise users. The framework enables developers and organizations to define repeatable AI workflows using structured files that teach assistants how to perform specific tasks consistently.

The company first introduced Agent Skills in October and is now positioning the technology as shared infrastructure for AI agents. The specification and reference SDK are hosted at agentskills.io, allowing developers to create portable skills that can be used across different AI systems. Skills are accessible across Claude.ai, Claude Code, the Claude Agent SDK, and the API at no additional cost.

Anthropic also introduced new administrative tools for Team and Enterprise customers, enabling centralized management of skill provisioning and default settings across organizations. The update includes a preview feature showing the full contents of a skill package before deployment.

Alongside the open standard, Anthropic launched a directory of partner-built skills from companies such as Atlassian, Figma, Canva, Stripe, Notion, Cloudflare, and Zapier. These integrations connect Claude with widely used enterprise applications, allowing users to automate tasks like creating reports, managing design assets, or updating project trackers.

The open standard approach follows Anthropic’s earlier release of the Model Context Protocol, which became widely adopted across the AI industry. The company said the Skills framework complements MCP by combining procedural knowledge with secure connectivity to external tools and data.