🧭 Agent Skills Goes Enterprise — Organisation Management and an Open Standard
Anthropic has significantly expanded Agent Skills — the framework for equipping Claude with reusable, composable capabilities — with two major additions: organisation-wide skill management for Team and Enterprise plan administrators, and the publication of an open standard specification for cross-platform skill portability. The open standard, published at agentskills.io, defines a vendor-neutral format for packaging and sharing skills, enabling the same skill to run on Claude.ai, in third-party Claude integrations, and on other platforms that adopt the specification. Microsoft has already announced adoption of the spec in VS Code and GitHub Copilot on the same day.
What's new for enterprise admins
- Org-level skill library: Team and Enterprise admins can now publish approved skills to an org-wide library, ensuring all employees access consistent, vetted capabilities rather than building ad-hoc workarounds.
- Usage analytics: Skill usage dashboards show which skills are being invoked most frequently, helping admins prioritise maintenance and identify gaps in the library.
- Partner-built skills directory: A curated directory of skills built by Anthropic's integration partners (including Salesforce, Workday, and Atlassian) is now accessible directly from the admin console.
- Permissioning: Skills can be scoped to specific teams, projects, or departments — preventing a skills designed for the finance team from surfacing in engineering contexts.
For developers building skills
Publishing to the open standard means your skill can reach users beyond the Claude.ai platform. Read the agentskills.io specification and structure your skill's manifest file accordingly — the schema is straightforward JSON and well-documented.
Agent Skills
enterprise
open standard
organisation management
retrospective
🧭 Claude in Chrome Now Available to All Paid Subscribers
Anthropic has opened the Claude Chrome extension to all paid Claude.ai subscribers — Pro, Team, and Enterprise — removing the earlier restriction that limited it to Max ($200/month) plan holders. The update coincides with new capabilities: workflow recording (teach Claude a repetitive browser task by demonstrating it once), Claude Code integration for terminal-to-browser testing workflows, and enterprise admin controls for managing which domains the extension is permitted to operate on. The extension uses Claude's Computer Use capability to read the browser state and take actions within the visible tab.
New capabilities in this release
- Workflow recording: Demonstrate a browser action (filling a form, navigating a workflow, copying data between tabs) and Claude records and can replay it. Useful for repetitive data-entry or research tasks.
- Terminal integration: Developers running Claude Code in a terminal can now pass browser context (the current page content, screenshots) to the terminal session and receive results rendered back in the browser.
- Admin controls (Team/Enterprise): Admins can specify approved domains, block Claude from operating on sensitive internal tools, and require confirmation prompts before Claude performs actions on specific sites.
Privacy note
The extension only reads content from the active tab, and only when explicitly invoked. It does not run in the background or send page content to Anthropic without user initiation. Enterprise plans can further restrict which domains the extension may read.
Chrome extension
computer use
workflow recording
Pro plan
retrospective
