Bootstrap a new Claude Code skill or command for the project. Use when creating a new skill from scratch — setting up the directory, frontmatter, required sections, and initial SKILL.md content. For evaluating, benchmarking, or iteratively improving an existing skill, use the skill-iterate skill instead.
Skills live in skills/<name>/SKILL.md (model-invoked) or commands/<name>/SKILL.md (user-invoked).
IMPORTANT: See the Filesystem Scope section in
CLAUDE.md.
All supported frontmatter fields:
| Field | Required | Description |
|---|---|---|
name | Yes | Kebab-case name — must match the directory name |
description | Yes | When/why to invoke this skill; drives auto-invocation matching |
user-invokable |
| No |
true to allow /plugin:name slash command invocation (default: false) |
tools | No | Comma-separated tool allowlist for agent definitions (omit to inherit defaults) |
model | No | Model override: opus, sonnet, or haiku (omit to inherit from parent) |
disable-model-invocation | No | true to prevent Claude from auto-invoking; use for user-only commands with side effects |
Notes:
tools and model are primarily used in agent files (agents/*.md), not skill filesdisable-model-invocation: true is appropriate for commands like /lead and /create-ticket that should only run when explicitly invokedskills/ that are also user-invokable should set user-invokable: true; commands in commands/ are always user-invokable and do not need this field| Type | Directory | Invocation |
|---|---|---|
| User-invoked slash command (side effects) | commands/ | /plugin:name only |
| Model-invoked background knowledge | skills/ | Claude auto-invokes based on context |
| Both user and model | skills/ | Either can invoke |
New SKILL.md files must include a ## Filesystem Scope section with the single-line reference:
## Filesystem Scope
> **IMPORTANT:** See the **Filesystem Scope** section in `CLAUDE.md`.
A realistic mini-skill showing all frontmatter fields and required sections:
---