Anthropic adds Agent Skills to Claude to package repeatable expertise
Oct 16, 2025 — Anthropic introduced Agent Skills, a portable format for packaging instructions, scripts, and resources that Claude can load when a task requires specialized behavior. Skills are presented as folders that Claude scans for relevance during a session, loading only the minimal files needed to complete the task.
What Skills are and how they behave
Skills are designed to make Claude more reliable on domain-specific work. Key properties highlighted by Anthropic:
- Composable: multiple skills can be stacked; Claude automatically identifies which skills are relevant and coordinates their use.
- Portable: the same skill format works across Claude apps, Claude Code, and the API.
- Efficient: skills load only the necessary content when required, which helps keep responses fast.
- Powerful: skills can include executable code, allowing tasks that benefit from deterministic computation rather than only token generation.
Claude searches the available skill set for relevant matches while working on tasks and invokes them without manual selection. Anthropic also provides a built-in skill-creator skill that guides the creation process by generating folder structures and formatting the SKILL.md file, reducing the need for manual file editing.
For a technical deep dive into the design pattern and architecture, see Anthropic’s engineering post: Equipping agents for the real world with Agent Skills.
Where Skills are available
Claude’s Skills functionality is available across Anthropic’s product lineup:
Claude apps
Skills are available to Pro, Max, Team, and Enterprise plans. Skills for common tasks are provided, with examples that can be customized or replaced with custom skills. Skills can be enabled via Settings: claude.ai settings. Team and Enterprise admins must enable Skills organization-wide.
Claude Developer Platform (API)
Skills can be added to Messages API requests, and a new /v1/skills endpoint supports programmatic skill versioning and management. Running skills requires the Code Execution Tool beta, which provides the secure runtime for executable components. Documentation and a developer-oriented guide are available here:
- Agent Skills overview: docs.claude.com - Agent Skills
- Code Execution Tool info: docs.claude.com - Code Execution Tool
- API skills guide: docs.claude.com - API skills guide
Developers can manage skill versions through the Claude Console and leverage Anthropic-provided skills for generating Excel spreadsheets, PowerPoint, Word documents, and fillable PDFs.
Claude Code
Skills extend Claude Code by integrating team workflows and expertise. Installable via plugins from the anthropics/skills marketplace, skills are automatically loaded when relevant. Manual installation is also supported by adding skills to ~/.claude/skills. The Claude Agent SDK includes the same Agent Skills support for building custom agents. See Claude Code docs: docs.claude.com - Claude Code skills and the example skills repo: github.com/anthropics/skills.
Getting started and safety notes
Anthropic links several resources for onboarding:
- User Guide and Help Center: Teach Claude your way of working using Skills and What are Skills
- Developer documentation and Anthropic Academy: Anthropic Academy
Because skills can execute code, Anthropic advises caution and recommends using skills from trusted sources. More on safe use is available in the support article: Using Skills in Claude — safety guidance.
For access to the original announcement and additional links, see Anthropic’s post: https://www.anthropic.com/news/skills
