Official API Reference

Official API Reference

Current frontmatter fields and constraints from official Claude Code documentation.

Required Fields

name

Type: String Required: Yes Constraints:

• Maximum 64 characters
• Lowercase letters, numbers, and hyphens only: [a-z0-9-]
• No XML tags
• No reserved words: "anthropic", "claude"

Examples:

name: pdf-processing       # Good
name: commit-helper        # Good
name: BigQuery-Analysis    # Bad - uppercase
name: claude-helper        # Bad - reserved word
name: my_skill             # Bad - underscores

description

Type: String Required: Yes Constraints:

• Maximum 1024 characters
• Non-empty
• No XML tags
• Must be third person

Template:

description: >-
  [What it does - actions, capabilities].
  Use when [trigger phrases, contexts, file types, user intents].

Examples:

# Good - specific, third person, trigger phrases
description: >-
  Extract text and tables from PDF files, fill forms, merge documents.
  Use when working with PDF files or when user mentions PDFs, forms,
  or document extraction.

# Good - action + context
description: >-
  Generate descriptive commit messages by analysing git diffs.
  Use when user asks for help writing commit messages or reviewing
  staged changes.

# Bad - first person
description: I can help you process Excel files

# Bad - vague
description: Helps with documents

# Bad - too long (over 1024 chars)
description: [very long description...]

Optional Fields

allowed-tools

Type: String (comma-separated) or Array Purpose: Limit which tools Claude can use when skill is active Default: All tools available

Examples:

# Comma-separated string
allowed-tools: Read, Grep, Glob

# YAML array
allowed-tools:
  - Read
  - Grep
  - Glob
  - Bash(git:*)

Tool patterns:

• Read - Specific tool
• Bash(git:*) - Bash limited to git commands
• Bash(npm:*) - Bash limited to npm commands
• mcp__server__tool - Specific MCP tool

model

Type: String Purpose: Override model for this skill's execution Default: Inherits from conversation

Examples:

model: claude-sonnet-4-20250514
model: claude-opus-4-20250514
model: inherit  # Explicit inherit

context

Type: String Purpose: Run skill in isolated context Values: fork

Example:

context: fork

Effect: Skill runs in separate context window, isolated from main conversation history.

agent

Type: String Purpose: Specify subagent type when using context: fork Values: Explore, Plan, general-purpose, or custom agent name

Example:

context: fork
agent: Explore

hooks

Type: Object Purpose: Define lifecycle hooks for the skill Events: PreToolUse, PostToolUse, Stop

Example:

hooks:
  PreToolUse:
    - matcher: Write
      prompt: |
        Before writing, verify the file path is within the project directory.
        If not, ask user for confirmation.

user-invocable

Type: Boolean Purpose: Control visibility in slash command menu Default: true

Examples:

user-invocable: false  # Hidden from menu
user-invocable: true   # Visible in menu (default)

Note: When false, skill is hidden from /skill-name menu but Claude can still:

• Invoke via Skill tool programmatically
• Discover and apply automatically based on description

disable-model-invocation

Type: Boolean Purpose: Prevent Claude from programmatically invoking via Skill tool Default: false

Examples:

disable-model-invocation: true   # Only user can invoke via /skill-name
disable-model-invocation: false  # Claude can also invoke (default)

Visibility Control Matrix

Complete Frontmatter Example

---
name: advanced-pdf-processing
description: >-
  Extract text, tables, and images from PDF files. Fill PDF forms
  programmatically. Merge and split PDF documents. Use when working
  with PDF files, document extraction, form filling, or PDF manipulation.
allowed-tools:
  - Read
  - Write
  - Bash(pdftotext:*)
  - Bash(pdftk:*)
model: inherit
user-invocable: true
hooks:
  PreToolUse:
    - matcher: Write
      prompt: Verify output path ends with .pdf before writing.
---

Skills in Subagents

Inject skills into subagent context:

---
name: code-reviewer
description: Review code for quality and security issues.
skills: security-check, style-guide
---

[System prompt for code-reviewer agent]

Effect: Full content of security-check and style-guide skills is injected into the subagent's context.

File Locations

Personal Skills (all projects)

~/.claude/skills/skill-name/SKILL.md

Project Skills (shared via git)

.claude/skills/skill-name/SKILL.md

Plugin Skills

plugin-name/skills/skill-name/SKILL.md

Validation Rules Summary

Troubleshooting

Skill not triggering

• Description too vague
• Missing trigger phrases
• Conflicting with another skill's description

Skill not loading

• SKILL.md not found (case-sensitive)
• Invalid YAML syntax (check --- delimiters)
• Name contains invalid characters

Skill errors during execution

• Missing dependencies in scripts
• Permission issues (chmod +x)
• Windows paths instead of forward slashes

Debug mode

claude --debug

Shows skill loading and selection decisions.