Agent Skills
An open, lightweight specification for packaging AI agent capabilities as portable skill folders, containing instructions plus optional scripts, references and templates.
Agent Skills is a standardized, open format for extending AI agent capabilities with specialized knowledge and workflows. A skill is a folder containing a SKILL.md file with metadata such as a name and description plus instructions for a specific task, and can bundle scripts, reference material, templates and other resources.
Skills package procedural knowledge and company-, team- or user-specific context into portable, version-controlled folders that agents load on demand, so the same skill can be reused across any compatible agent. Agents load skills through progressive disclosure in three stages: discovery, where only a skill's name and description load at startup; activation, where the agent reads the full SKILL.md once a task matches; and execution, where the agent follows the instructions and optionally runs bundled code or loads referenced files.
The format was originally developed by Anthropic. It is written in Python and licensed Apache-2.0.