Beta
AI Foundry is in beta. We are actively shaping the product, so things may change as we iterate. Your feedback is welcome.
Skill
A Skill is a catalog resource that encapsulates a reusable, higher-level AI capability. Where a Tool wraps a single atomic operation (call this API, run this query), a skill captures a multi-step capability along with the knowledge, templates, scripts, and assets needed to exercise it, for example "summarise a document", "classify intent", or "draft a reply in brand voice".
Skills can be referenced both by Agents or Playbooks. They act as a reusable building block that can be shared across multiple agents without duplicating the underlying logic.

Anatomy of a skill
A skill bundles four optional content sections alongside its metadata:
manifest: the primary Markdown documentation of the skill (SKILL.md), covering what it does, how to invoke it, and optionally the information about other elements listed below.refs: named reference materials the skill draws on (e.g. style guides, domain glossaries, external links).assets: named templates or structured content the skill produces or consumes (e.g. a response template, a classification taxonomy).scripts: named code snippets or pseudocode that implement part of the skill logic.
Skill schema
| Field | Required | Description |
|---|---|---|
Title | Yes | Display name shown in the UI. |
Name | Yes | Unique identifier. Immutable after creation. Referenced in Agent.spec.skills and Playbook.spec.skills. |
Description | No | Short description of what the skill enables. |
Manifest | Yes | Markdown documentation of the skill. Must be non-empty. Rendered in the detail view with a source/preview toggle. |
Refs | No | A key-value map where each key is a reference name and each value is the reference content (Markdown). |
Assets | No | A key-value map where each key is an asset name and each value is the asset content (templates, schemas, etc.). |
Scripts | No | A key-value map where each key is a script name and each value is the script content (pseudocode, shell, Python, etc.). |
Skills vs. Tools
| Aspect | Skill | Tool |
|---|---|---|
| Abstraction level | High: captures a multi-step, knowledge-rich capability | Low: wraps a single external operation |
| Implementation location | Documented in the catalog (manifest, scripts, assets) | External service or MCP server |
| Content stored in catalog | Yes (manifest, refs, assets, scripts) | Yes |
| Typical use | Writing, reasoning, classification patterns | API calls, database queries, code execution |