Product Categories

Choose the right product category for your MyClaude product. Each of the 9 categories has specific requirements and file structures.

Category determines how the CLI installs your product, where files land in the buyer's environment, and which validation rules apply at publish time. Pick by answering: what does a buyer do immediately after installing this?

Category overview

Skill

A skill is a callable prompt routine. Buyers invoke it with /skill-name inside any Claude Code session. It is the most common product type — specific, reusable, and instantly useful.

Required files:

my-skill/
  SKILL.md      # Required — prompt instructions
  vault.yaml    # Required — manifest
  README.md     # Optional — buyer-facing docs

Use when: Your product does one thing well and is invoked on demand. Code reviewers, commit message generators, security auditors, documentation writers.

Not when: Your product needs multiple coordinated agents (use Squad), or it is a full application (use Application).

Guide: Writing Skills Guide

Squad

A squad is a team of agents with assigned roles, coordinated to handle complex tasks that benefit from specialization. The SQUAD.md defines the roster, roles, and coordination protocol.

Required files:

my-squad/
  SQUAD.md          # Required — roster, roles, coordination rules
  agents/
    agent-name.md   # One file per agent in the squad
  vault.yaml

Use when: The task is too complex for one agent — different expertise is needed at different stages. Content teams, engineering squads, research teams.

Not when: One well-written skill handles the use case. Squads are higher overhead for buyers to operate.

Agent

An agent is a single-purpose AI entity with a defined knowledge domain and behavior. Agents are often assembled into squads, but can also be installed and used independently.

Required files:

my-agent/
  AGENT.md      # Required — identity, capabilities, constraints
  vault.yaml

Use when: You are building a specialist that will be composed into larger systems. A TypeScript reviewer, a security auditor, a brand voice enforcer.

Not when: Your product is invoked directly by the buyer as a complete tool — use Skill.

Workflow

A workflow is a multi-step automation pipeline. It chains skills, agents, and external tools into a repeatable process with defined inputs, steps, and outputs.

Required files:

my-workflow/
  WORKFLOW.md       # Required — steps, inputs, outputs, coordination
  steps/
    step-name.md    # Optional — step-specific instructions
  vault.yaml

Use when: The value is in the sequence, not any single step. CI/CD pipelines, deploy workflows, multi-stage analysis.

Not when: The process has one or two steps — a skill handles that more simply.

Design System

A design system product delivers a complete token set, component library definition, and design rules. The DESIGN-SYSTEM.md gives Claude enough context to generate consistent UI code within the system's constraints.

Required files:

my-design-system/
  DESIGN-SYSTEM.md    # Required — tokens, rules, component patterns
  tokens/
    colors.yaml       # Optional — structured token files
    typography.yaml
  vault.yaml

Use when: You have a complete, coherent design language with documented tokens and rules. Not when you have a color palette and nothing else.

Prompt

A prompt is a reusable system prompt or template. Buyers copy it into their Claude.ai system prompt, their project CLAUDE.md, or use it as a task template. Prompts are the simplest product type — no installation ceremony.

Required files:

my-prompt/
  PROMPT.md     # Required — the prompt content
  vault.yaml

Use when: The value is in the prompt itself. Specialized analyzers, tone/style prompts, domain-specific system prompts that buyers adapt to their projects.

CLAUDE.md

A CLAUDE.md product is a project configuration file with behavioral rules, context, and directives for Claude Code. It installs into the buyer's project root, shaping how Claude behaves throughout that project.

Required files:

my-claude-md/
  CLAUDE.md     # Required — the configuration
  vault.yaml

Use when: You have a well-tested Claude Code configuration for a specific context — enterprise compliance rules, framework-specific conventions, team coding standards.

Install behavior: Unlike other categories, claude-md installs to the buyer's current project root, not a central directory. Buyers run myclaude install from inside their project.

Application

An application is a complete, deployable product built for or with Claude Code. It may include scaffolding scripts, multi-file structures, and configuration for external services.

Required files:

my-app/
  APP.md            # Required — overview, setup, usage
  src/              # Optional — application source files
  setup.sh          # Optional — setup script
  vault.yaml

Use when: You are selling a complete system, not a component. Dashboard generators, API scaffolders, full-stack starters.

Min CLI version: Applications often use newer CLI features. Set min_cli_version in vault.yaml when your install process requires a specific CLI version.

Choosing the right category

Answer these questions in order:

1. Does the buyer invoke it on demand with a slash command? → Skill
2. Does it need multiple specialized agents? → Squad (or individual Agents for the components)
3. Is it a multi-step automation? → Workflow
4. Is it a design language with tokens and rules? → Design System
5. Is it a system prompt or reusable template? → Prompt
6. Is it a project-level Claude Code configuration? → CLAUDE.md
7. Is it a complete, deployable application? → Application

When in doubt, start with Skill. It is the simplest structure and the easiest for buyers to adopt. Complexity can always be added in v2.

Related pages

• Writing Skills Guide — deep dive into the most common category
• vault.yaml Specification — category-specific field behavior
• Publishing Your First Product — publish any category in 5 minutes