If you have spent any time with Claude, you know the drill: you explain the context, set the tone, paste the guidelines, and then finally get to the actual task. Every session. Every time.
Skills is Anthropic’s attempt to fix that.
Launched in beta in October 2025 and expanded into an open standard in December, Skills lets developers and organisations package repeatable instructions, context, and resources into reusable folders that Claude loads automatically when they become relevant. Instead of starting from scratch each session, Claude picks up the playbook on its own.
What Skills Actually Is
At the core, a Skill is a folder. Inside that folder lives a SKILL.md file with instructions, any supporting scripts, and whatever reference material Claude needs to do the job right. When a task matches what a Skill covers, Claude loads it. When it doesn’t, Claude ignores it. There’s no manual invocation required, though developers can also trigger Skills directly with slash commands.
Anthropic-managed Skills come pre-built for working with common file formats: Excel spreadsheets, PowerPoint presentations, Word documents, and fillable PDFs. These are the Skills already running quietly in the background when Claude helps you build a slide deck or format a spreadsheet inside claude.ai.
Developers can also build their own Skills through a /v1/skills endpoint, version and manage them in the Claude console, and integrate them into existing workflows. A team could build a Skill that pulls structured data from an internal database, another that formats meeting summaries in a specific style, and another that creates Jira tickets with the right fields already filled in.
Why It Matters Beyond the Technical Details
Skills are reusable instruction sets that teach Claude specific workflows, standards, and domain knowledge: brand style guidelines, email templates, task creation in tools like Jira and Asana. For teams that use Claude daily, that means not re-explaining house rules every single time.
The bigger development came in December 2025, when Anthropic published Agent Skills as an open standard. Skills built in Claude can be used in other models and platforms, like ChatGPT or Cursor, that adopt the standard. In practice, this means a Skill you build today is not locked to Claude. It travels with you.
The same SKILL.md format already works across Claude Code, Cursor, Gemini CLI, and other compatible AI agents. The push toward portability is deliberate: Anthropic is positioning Skills as infrastructure, not just a product feature.
How Developers Are Using It
Enterprise teams at companies like Zapier, TELUS, and Bridgewater have already adopted Skills at scale. Use cases range from code review workflows to automated report generation to internal documentation formatting.
A developer could create a Skill to fetch structured data from a company database, compose personalised email responses using CRM data, summarise meeting transcripts in a specific format, or trigger actions in third-party applications such as Slack or Notion.
For Claude Code users specifically, Skills can be installed via plugins from the anthropics/skills marketplace, with Claude loading them automatically when relevant, and shared across teams through version control.
The most recent updates also bring Skills support to the Claude for Excel and Claude for PowerPoint add-ins, along with a new Customize section in Claude Desktop that groups skills, plugins, and connectors in one place.
The Catch
Skills execute code. That is what makes them powerful, and also what requires some caution. Anthropic has been upfront: stick to trusted sources. A Skill from an unknown third party has the same level of access to your environment as one you built yourself.
Each Skill runs within clearly defined boundaries, ensuring Claude only accesses data and executes actions explicitly allowed by the developer. But those boundaries are only as good as how carefully the Skill was written.
For organisations considering enterprise-wide deployment, the advice is consistent: start with one workflow, get familiar with how a Skill behaves, then expand from there.
The Bottom Line
Skills is not a flashy announcement. It is the kind of infrastructure change that compounds quietly. Teams that invest time building their own Skills now, for their brand guidelines, their internal tools, their specific output formats, will interact with Claude differently than teams that don’t.
The open standard element is what makes this worth paying attention to beyond Claude itself. If Skills becomes the way AI agents share institutional knowledge across platforms, the teams building that knowledge library today have a head start.
and then