Core Concepts

Skills & Tools

Skills are the building blocks of SkillRouter. Learn how they work, how they're organized, and how to use them effectively.

What is a Skill?

A skill in SkillRouter is a discrete, executable capability that an AI agent can use to interact with the external world. Skills encapsulate:

  • API Integration: Connection to a third-party service (like Stripe, GitHub, or SendGrid)
  • Authentication: Managed credentials and OAuth flows handled by SkillRouter
  • Input Schema: Defined parameters the skill accepts, with validation
  • Output Format: Structured response format for consistent agent consumption
  • Error Handling: Built-in retry logic and graceful error responses

Skills vs. Tools

While the terms are often used interchangeably, we make a distinction:

Tools

Raw API integrations with basic functionality. Think of tools as the underlying capabilities.

Example: "HTTP Request to Stripe API"

Skills

Curated, optimized tools with semantic descriptions, managed auth, and built-in best practices.

Example: "Process a payment with Stripe Checkout"

Skill Properties

Every skill in the registry has the following properties:

PropertyDescription
idUnique identifier (e.g., `stripe-checkout`)
nameHuman-readable name
descriptionSemantic description for discovery
providerThe service provider (e.g., Stripe)
categorySkill category for organization
input_schemaJSON Schema defining required parameters
output_schemaJSON Schema defining response format

Skill Health & Metrics

Every skill is continuously monitored for health and performance:

Success Rate

Percentage of executions completing without error

Latency

Average and p99 response times

Usage

Total executions and active agents