Documentation / Agent skills / Introduction

Skills: what is a skill?

Understand skills as portable units of procedure and domain knowledge—the building blocks that let agents do real work beyond plain text generation.

← Documentation overviewAgent skills hub

A skill is a portable, self-contained unit of domain knowledge and procedural logic that enables an AI agent to perform specific tasks. Skills embody the know-how—multi-step reasoning, orchestration, decision rules, sequencing logic, validation steps, conditional branching, and output formatting standards.

Why skills matter

Skills are the building blocks of agentic systems. Without skills, an AI is just a language model—it can generate text, but it cannot do anything. Skills give agents:

  • Autonomy — The ability to execute tasks without step-by-step human guidance.
  • Domain expertise — Encoded knowledge and procedures for specific domains.
  • Reusability — Skills can be shared across different agents and use cases.
  • Composability — Skills can be chained together to form complex workflows.

How skills work in Agentic Hosting

In the Agentic Hosting platform, skills are:

  1. Loaded on demand — Skills are only loaded when needed, keeping the agent lean.
  2. Versioned — Each skill has a version number for tracking changes.
  3. Categorized — Skills are organized by category (for example automation, research, ops).
  4. Tagged — Tags enable discovery and filtering of skills.

Skill anatomy

A skill consists of:

ComponentDescription
MetadataYAML frontmatter with title, slug, version, author, tags, category, status.
OverviewBrief summary of what the skill does.
UsageExample invocation and parameters.
ImplementationTechnical details, inputs, outputs, edge cases.
ExamplesReal-world usage examples.
Related skillsLinks to other related skills.

What skills can do

Skills can encapsulate virtually any procedural or knowledge-based task:

  • Data operations — Extract, transform, load, validate data.
  • API integrations — Connect to external services and APIs.
  • Workflow orchestration — Chain multiple steps into a process.
  • Decision making — Apply business logic and conditional rules.
  • Analysis — Process and analyze data, generate insights.
  • Automation — Automate repetitive tasks and processes.
  • Research — Gather, synthesize, and report on information.

Next steps