Skills and plugins are packaged capabilities that extend what an AI system can do.
They may combine instructions, tools, resources, configuration, and workflow logic into a reusable capability.
A skill packages capability. A tool performs a specific action.
A smartphone app.
The phone → AI system
Individual device functions → Tools
Installed app → Skill / Plugin
camera · location · network · storage
into one useful capability.
This analogy is for understanding only. Different AI platforms define “skill” and “plugin” differently.
“Prepare a daily sales report.”
Database Query
Calculator
File Creation
One packaged skill can orchestrate multiple tools.
search_web()
query_database()
send_email()
May combine instructions, multiple tools, workflow, and resources.
Tool = individual action
Skill / Plugin = reusable packaged capability
Defines a usable capability or workflow.
Defines a standardized way for AI applications to connect to external tools and resources.
Skill / Plugin describes capability packaging. MCP describes connection standardization.
Decides what actions to take toward a goal.
Agent → decides
Provides a capability the agent may use.
Skill → provides reusable capability
Different platforms may use terms such as:
Skill · Plugin · Extension · Action · App · Tool Package
The exact implementation varies by platform. These names are not a strict universal taxonomy.
The model generates or selects outputs; the package adds reusable capability around it.
A package may use tools, but a tool is one specific action.
A package describes capability; MCP standardizes connections.
An agent decides what to do; a skill provides something it can use.
A skill or plugin packages reusable capabilities that an AI system can use to complete tasks.