Skills

Skills follow the Agent Skills format and are packaged using the Agent Plugins layout.

Each skill is an immediate child of skills/ with its own SKILL.md. Nested directories can contain supporting files, but they are not discovered as additional skills.

Layout

skills/
  frontend-design/
    SKILL.md
    references/     # optional
    scripts/        # optional
    assets/         # optional

A skill can include any additional files it needs. references/, scripts/, and assets/ are common conventions.

SKILL.md starts with YAML frontmatter followed by the instructions:

---
name: frontend-design
description: Design distinctive UI with deliberate typography, color, and layout. Use when creating or substantially redesigning an interface.
---

# Frontend Design

…

name must match the skill directory name.

description should explain both what the skill does and when it should be used. The agent sees skill names and descriptions during discovery and loads the full SKILL.md only when the skill is activated.

Invocation in Reforma

By default, a skill can be activated by the model when relevant and invoked directly by the user.

Reforma adds two optional frontmatter fields to control invocation:

For a user-only command:

---
name: audit-ui
description: Audit the current interface against the UI guidelines and report problems.
disable-model-invocation: true
---

The user can invoke it directly, for example with /audit-ui, but the model will not select it automatically.

For an agent-only skill:

---
name: install-components
description: Install missing UI components from the configured component registry when implementation requires them.
disable-user-invocation: true
---

The model can activate it when relevant, but it is hidden from user invocation.

These invocation fields are Reforma extensions; they are not part of the portable Agent Skills specification.

Skills vs rules

Use a skill for knowledge or a workflow that should be loaded on demand.

Use a rule for instructions that should apply on every turn.